Anoctamin Calcium-Activated Chloride Channels May Modulate Inhibitory Transmission in the Cerebellar Cortex

Calcium-activated chloride channels of the anoctamin (alias TMEM16) protein family fulfill critical functions in epithelial fluid transport, smooth muscle contraction and sensory signal processing. Little is known, however, about their contribution to information processing in the central nervous system. Here we examined the recent finding that a calcium-dependent chloride conductance impacts on GABAergic synaptic inhibition in Purkinje cells of the cerebellum. We asked whether anoctamin channels may underlie this chloride conductance. We identified two anoctamin channel proteins, ANO1 and ANO2, in the cerebellar cortex. ANO1 was expressed in inhibitory interneurons of the molecular layer and the granule cell layer. Both channels were expressed in Purkinje cells but, while ANO1 appeared to be retained in the cell body, ANO2 was targeted to the dendritic tree. Functional studies confirmed that ANO2 was involved in a calcium-dependent mode of ionic plasticity that reduces the efficacy of GABAergic synapses. ANO2 channels attenuated GABAergic transmission by increasing the postsynaptic chloride concentration, hence reducing the driving force for chloride influx. Our data suggest that ANO2 channels are involved in a Ca2+-dependent regulation of synaptic weight in GABAergic inhibition. Thus, in balance with the chloride extrusion mechanism via the co-transporter KCC2, ANO2 appears to regulate ionic plasticity in the cerebellum.


Introduction
Calcium-activated chloride channels of the anoctamin (alias TMEM16) family of membrane proteins provide a chloride conductance that operates under the control of intracellular Ca 2+ signals (recent review: [1]). Many different cell types express anoctamin proteins. The anoctamin chloride channels anoctamin 1 (ANO1, TMEM16A) and anoctamin 2 (ANO2, TMEM16B) have been established as Ca 2+ -activated Clchannels with defined physiological functions [2][3][4]. They are involved in epithelial Cltransport, smooth muscle contraction and neuronal signal processing. Anoctamin channels show highly polarized expression patterns in performed using the Vectastain ABC kit guinea pig IgG (biotinylated secondary antibody dilution 1:200; PK-4007; Vector Labs). Sections were incubated in 0.5% H 2 O 2 for 1 hour and washed 3 × 5 minutes with PBS, followed by incubation in blocking solution (5% normal goat serum, Sigma-Aldrich, G9023, in PBS with 0.5% Triton X-100, 0.05% NaN 3 ) for 1 hour. The primary antiserum was diluted in blocking solution and applied overnight at 4°C, then washed for 3 × 10 minutes with PBS. Sections were then incubated in biotinylated secondary antibodies, diluted in blocking solution, for 2 hours, washed 3 × 10 minutes in PBS, incubated in ABC solution (from the ABC Kit, one drop of A and one drop of B in 20 ml PBS, prepared at least 30 minutes before use) for 1 hour, and washed 3 × 10 minutes with PBS. Sections were developed for 5 minutes with DAB-H 2 O 2 solution (1μl of 1% H 2 O 2 in 1 ml 3,3'-diaminobenzidin solution; one pellet DAB in 10 ml PBS; Sigma-Aldrich D5905) to start the reaction, washed with PBS intensively, and mounted on glass slides using Aqua-Poly/Mount (Polysciences, Inc., 18606). For immunofluorescence staining, sections were incubated by blocking serum for 1 hour, and the primary antisera (diluted in blocking solution) were applied overnight at 4°C for brains, and 2 hours at room temperature for noses. After washing 3 x 10 min with PBS, the secondary antisera, conjugated with Alexa Fluor tags (Molecular Probes, Inc.), were incubated for 2 hours, washed 3 x 10 minutes with PBS. To visualize cell nuclei, the slices were incubated in DAPI (0.3 μM 4,6-diamidin-2-phenylindol in PBS; Sigma-Aldrich 32670) for 3 minutes, washed 3 x 5 minutes in PBS and mounted on glass slides with Aqua-Poly/Mount. To stain all neurons, we incubated sections in NeuroTrace 1 530/615 (Invitrogen N21482, dilution 1:500; Life Technologies, Inc.) for 5 minutes before the DAPI treatment, followed by 3 x 5 minutes wash with PBS. Images were obtained using a Nikon C1 confocal microscope. All imaging data presented are single-plane images, no Z-stacks are used in this paper.

Expression and molecular characterization of cerebellar Ano1 and Ano2
Semiquantitative RT-PCR analysis was performed on total RNA of cerebellum and, for comparison, on RNA from olfactory epithelium. RNA was isolated using the MagJet RNA Kit (Thermo Scientific). cDNA was synthesized using 5 μg total RNA, oligo (dT) 18 primer, and the Maxima H Minus First Strand cDNA Synthesis Kit (Thermo Scientific). PCR amplification was performed on 1 μl (250 ng) single-stranded cDNA with DreamTaq PCR Master Mix (Thermo Scientific) using the primer pairs ANO1/F861 and ANO2/F865 (Table 1). Cycling conditions were 95°C for 3 min, followed by 28 to 34 cycles at, respectively, 95°C for 30 s, 62°C for 30 s, 72°C for 60 s, and finally 72°C for 8 min. The bands were resolved by gel electrophoresis and were verified by sequencing after purification by GeneJET Gel Extraction Kit (Thermo Scientific). To characterize the splice variants of the Ano1 and Ano2 transcripts, PCR was performed on cerebellar cDNA using the primers listed in Table 1. For Ano1, we used a set of 5 primer pairs which covered the entire transcript according to Genbank Acc. No. NM_178642.5. The forward primer of pair ANO1/F581 matched to the sequence at an alternative predicted translation start site (position 98 in NM_178642.5) and ANO1/F857 matched the sequence at the translation start site of isoform a (position 269). ANO1/F855, ANO1/F847 and ANO1/F845 consecutively matched the following sequence of the open reading frame. The primer pair ANO1/F581 resulted in no product while the four other primer pairs resulted in abundant PCR products of predicted size. By sequencing the PCR products, we found that the ANO1ac variant is expressed in the cerebellum. For ANO2, the primer pairs ANO2/F871, ANO2/F870 and ANO2/F869 consecutively matched the sequences of three alternative predicted translation start sites according to Genbank Acc. No. NM_153589.2. ANO2/F867, while By sequencing all PCR products, we found that the cerebellar ANO2 isoform is identical to the main olfactory isoform [6,9].

Immunoblot analysis
Main olfactory epithelium, eyes, olfactory bulb and cerebellar tissue were dissected from wildtype C57BL/6 (Black6) or Ano2 -/mice. A whole-protein extraction method was used for most of the samples. For biochemical assays of ANO2 expression in olfactory bulb and cerebellum, the Qproteome Cell Compartment Kit (Qiagen) was used for protein extraction because it was reported to produce particularly high yield of membrane protein [35]. The supernatant of fraction 2, containing primarily membrane proteins, was separated by SDS-PAGE on 10% gels and electro-blotted to PVDF membranes (Machery & Nagel; Germany) using a semidry blotting apparatus. Membranes were blocked with 5% milk powder (in PBS / 0.1% Tween 20) for 1 hour and incubated with the primary antibodies overnight. The blots were washed three times with 0.1% Tween 20 in PBS and incubated for 1 hour with a horseradish peroxidase-conjugated secondary antibody. The blots were washed again, and the ECL plus enhanced chemoluminescence system (GE Healthcare, Germany) was used to monitor bound antibodies. Antibodies used for immunoblotting were guinea-pig anti-ANO1 (dilution 1:400; C-terminus encoding amino acids 962-1040; marked "ANO1 in " in the text) [33], and a rabbit ANO2 antiserum directed against the extracellular loop that connects TMDs 5 and 6 in ANO2 (Alomone Labs, ACL012, dilution 1:1000, marked "ANO2 ex " in the text). The specificity of the ANO2 ex antiserum was verified using olfactory epithelium and VNO, comparing wild-type and ANO2 -/mice. Furthermore, we used goat anti-actin (Santa Cruz sc-1615, dilution 1:1000), rabbit anti guinea pig HRP secondary antibody (Sigma, dilution 1:20000), goat anti rabbit HRP secondary antibody (Sigma, dilution 1:30000), donkey anti goat HRP secondary antibody (Jackson ImmunoResearch, dilution 1:10000).

EYFP-tagged Ano1 and Ano2 expression plasmids
For heterologous expression in HEK 293 cells, EYFP-tagged mouse cerebellar ANO1 and ANO2 were used. For ANO1ac, the primer pair ANO1/F524 (Table 1) was used for full length cloning from cerebellar cDNA. For controls, the ANO1abc isoform was amplified from mouse Ano1abc pCMV-SPORT6 plasmid (kindly provided by Dr. Rainer Schreiber und Dr. Karl Kunzelmann, University of Regensburg) using the same primers. An EcoRI site was introduced at the 5'end and a BamHI site at the 3' end of the fragments for fusing EYFP to the C-terminus in the expression vector pEYFP-N1 (Takara Bio Europe/Clontech, France). The ANO2-pEYFP-N1 expression plasmid was kindly provided by Dr. Johannes Reisert (Monell Chemical Senses Center, Philadelphia).

Electrophysiology
The cerebellum was removed from the skull directly after sacrificing the animals, and was mounted with cyanoacrylate glue in a vibratome chamber (Leica VT1000S) filled with the respective artificial cerebrospinal fluid (ACSF) at 34-37°C. Sagittal tissue slices were cut at 220-250 μm and transferred to an incubation chamber filled with ACSF at 34°C for 30 min. After this, the incubation chamber was kept at room temperature. All solutions were saturated with 95% O 2 / 5% CO 2 [36]. For patch-clamp experiments, we transferred the tissue slices onto the stage of an upright, water immersion microscope (Nikon Eclipse E600FN). The recording chamber was continuously perfused with oxygenated ACSF at room temperature. Visual control was achieved with a camera system (Nikon DN100), and Purkinje cells were identified by their shape and location. To obtain whole-cell recordings, borosilicate glass micropipettes (Science Products, Hofheim, Germany; GB-150F-10) were made using a horizontal puller (Sutter Instruments; P97) to a resistance of 2-3 MO. Pipettes were filled with an intracellular solution and positioned on the Purkinje cell soma. After obtaining a tight seal, the plasma membrane inside the pipette tip was disrupted by suction to establish the whole-cell configuration. The drug perfusion system operated at 1-2 ml/min for a complete exchange of the bath solution.
To block presynaptic cannabinoid receptors, which tend to reduce inhibitory signals [37], the CB1 antagonist AM251 (2 μM) was included in the bath solution. Electrical signals were recorded using a patch-clamp amplifier (HEKA Electronics, Lambrecht, Germany; EPC-8) and the WinWCP (4.6.1) program provided by the University of Strathclyde, Glasgow, UK. Capacitance compensation was applied; series resistance was not compensated. Holding voltages were corrected for liquid junction potentials. Stimulation of climbing fibers was triggered by a constant current source (Digitimer, Welwyn Garden, UK; DS7A) through an ACSF-filled pipette. Postsynaptic currents were counted off-line before and after stimulation of climbing fibers. 3-minute traces were scanned by a peak-detection software (Mini-Analysis Program, Synaptosoft Inc. For heterologous expression, EYFP-tagged mouse cerebellar Ano1 or Ano2 was transfected into HEK 293 cells by Ca 2+ -phosphate co-precipitation. 24 h after transfection, expression was confirmed by yellow fluorescence, and cells were examined in whole-cell configuration at -70 mV. The bath solution contained (mM): 150 CsCl, 10 HEPES, 10 EGTA; pH 7.4 (CsOH); the pipette solutions contained 133.5 mM CsCl, 8.26 mM CaCl 2 , 10 mM HEDTA, 10 mM HEPES; pH 7.0 (CsOH), to give a free Ca 2+ concentration of 7.5 μM [38]. Currents were recorded immediately after whole-cell break-through as Ca 2+ diffused into the cell. To evaluate the data for ANO2, the maximal current amplitudes were related to the individual cell capacitance. The current densities from 19-30 cells were averaged for each test. With ANO1 expressing cells, we did not obtain individual values for cell capacitance because large chloride conductances prevented accurate determination. In this case, current densities were calculated using the average cell capacitance of ANO2-expressing HEK 293 cells in our experiments. To test the effect of T16Ainh-A01, 5 μM of the compound were continuously included in the bath solution. Statistical analysis was done using Student´s t test. Error bars indicate SEM; significance levels were p < 0.05 ( Ã ), p < 0.01 ( ÃÃ ), p < 0.001 ( ÃÃÃ ) or p< 0.0001 ( ÃÃÃÃ ).

Results
Detection of ANO1 and ANO2 expression in the mouse cerebellum ANO1 and ANO2 channels are thought to have similar membrane topology. Hydropathy analyses and various functional assays have pointed to a model with 8 transmembrane domains [39][40][41] which was refined by X-ray crystallography to a 10 TMD model ( Fig 1A) [42]. We compared the cerebellar expression levels of both genes with the olfactory neuroepithelium because the expression of both proteins is well characterized in that tissue [33]. The PCR signal for ANO1 cDNA was comparable between cerebellum and olfactory epithelium, while the ANO2 cDNA was weaker in cerebellum than in olfactory epithelium ( Fig 1B). Both for ANO1 channels and for ANO2 channels, several alternative translation forms exist [5,6,41]. For ANO1, the positions of four relevant exons are indicated as a-d in Fig 1A. To identify the cerebellar ANO1 isoform, we used a set of 5 overlapping primer pairs ( Table 1) that, together, covered the entire open reading frame of the transcript. By sequencing the PCR products, we found that the ANO1ac variant is expressed in the cerebellum. This isoform was previously shown to encode Clchannels with particularly high apparent Ca 2+ sensitivity (K D = 0.15 μM at -40 mV; [5]). ANO2 proteins exist in the two isoforms A and B in olfactory receptor neurons [6] (Fig 1A). By sequencing the PCR products from cerebellum, we found the isoform B to be the predominant ANO2 variant. Isoform B is also the prevalent form in olfactory receptor neurons. It may contain a regulatory motif at a position homologous to segment c of ANO 1. This motif is present in photoreceptors but not in olfactory receptor neurons [26]. We found that it is also absent from the cerebellar ANO2 sequence. In immunoblots from membrane-protein preparations, the ANO1 in serum labeled a band at the expected size of 115 kDa in olfactory Membrane topology model for anoctamin Ca 2+ -activated Clchannels based on the Xray structure of a fungal TMEM16 protein [42]. The transmembrane domains 5 and 6 are thought to provide the pore-lining region in the homodimeric channel [95]. Five negatively charged amino-acid residues (E, D) and an asparagine residue (N) in transmembrane domains 6-8 serve as Ca 2+ -binding sites involved in channel gating [39][40][41]. Four alternatively spliced segments (a-d) determine the apparent Ca 2+ -sensitivity of the ANO1 channel [5]. ANO2 has two isoforms A and B and a regulatory motif at a position homologous to segment c in ANO1 [6]. (B) RT-PCR analysis from mouse olfactory epithelium (OE) and mouse cerebellum (CB) yield similarly strong ANO1 signals in cerebellum but weaker signals for ANO2. (C) Immunoblots obtained from lysates of cerebellum (CB) and main olfactory epithelium (OE) from wild-type and Ano2 -/mice show an ANO1-specific signal at~120 kDa with the ANO1 in antiserum. (D) Rabbit anti-ANO2 ex serum stains ANO2-specific bands (asterisks) in immunoblots obtained from lysates of main olfactory epithelium (OE) and eye, as well as in membrane-protein preparations of main olfactory bulb (OB) and cerebellum (CB). ANO2 bands are not present in immunoblots from Ano2 -/mice.
doi:10.1371/journal.pone.0142160.g001 epithelium and in cerebellum, both in wild-type and in Ano2 -/mice ( Fig 1C). A weaker second band was visible at~95 kDa. For detection of ANO2 protein, we used the ANO2 ex antiserum which labeled a discrete band in cerebellum, as well as broader bands in preparations from olfactory epithelium, olfactory bulb and eye, all of which were absent in Ano2 -/mice ( Fig 1D). These signals are characteristic for the glycosylated ANO2 protein at~120 kDa in eye and at 150-170 kDa in olfactory tissues [8]. In our membrane-protein preparation, the cerebellar ANO2 protein appeared at~120 kD with no evidence for pronounced glycosylation. These data consistently demonstrate that ANO1 and ANO2 proteins are expressed in the mouse cerebellum.

Differential expression of ANO1 and ANO2 in the cerebellar cortex
Purkinje cells, the large output neurons of the cerebellar cortex, receive excitatory input from granule cells and climbing fibers, as well as inhibitory input from stellate cells, basket cells and Golgi cells (Fig 2A). To find out which cells express ANO1 and ANO2, cryosections were prepared without pre-fixation of the tissue. DAB-labeled ANO1 in antibodies stained the Purkinje cell layer as well as scattered cells in the molecular layer and granule cell layer (Fig 2B and 2C). Inhibitory interneurons in the granule cell layer ( Fig 2D) and in the molecular layer ( Fig 2E) were immunopositive both for ANO1 and for glutamate decarboxylase (GAD), a marker for GABAergic neurons. We used a mouse line that expressed Cre recombinase in all GAD67-expressing cells [32], and immunostained with an antiserum raised against Cre recombinase [34]. The labeled neurons in the granule cell layer are probably Golgi cells (Fig 2G) while the GAD/ ANO1-positive cells in the molecular layer are basket/stellate cells (Fig 2H). Thus, apparently all GABAergic cells in the cerebellar cortex were ANO1-positive. The somata of Purkinje cells also expressed ANO1 (Fig 2F). Co-staining with the neuronal marker NeuroTrace 1 revealed that all Purkinje cells contained ANO1. However, ANO1 immunosignals were faint in, or absent from, the Purkinje cell dendrites. The preadsorption control with the immunizing peptide showed no immunosignal (Fig 2I).
Immunostaining of cerebellar cortex cryosections with ANO2 in antiserum revealed a different expression pattern compared to ANO1. Like ANO1, the ANO2 immunosignal was absent from granule cells. However, ANO2 was also absent from the inhibitory interneurons. Instead, the protein was located in the Purkinje-cell dendrites (Fig 3A). Within the somata of Purkinjecells, ANO2 immunosignals appeared to be confined to the perinuclear area, presumably the rough endoplasmic reticulum. This contrasts with ANO1 expression, as ANO1 protein can be detected all through the soma, but not in the dendrites (Fig 3B). For ANO2, a viable knockout mouse line is available [8]. To test the specificity of the cerebellar ANO2 signal, we stained cryosections of Ano2 -/mice and detected no immunosignals under the same experimental conditions (Fig 3C). The only other structure in the brain reported to be immunopositive for ANO2 is the olfactory bulb where the axons of olfactory receptor neurons coalesce onto glomeruli to form their synapses [8]. Fig 3D and Fig 3E depict the ANO2 immunosignal in olfactory bulb as a positive control and its absence from the Ano2 -/mouse.
These data reveal a differential expression pattern of ANO1 and ANO2 in the cerebellar cortex. Only ANO2 is present at detectable density in the dendrites, the site of synaptic plasticity of inhibitory transmission as discovered by Satoh et al. (2013) [31]. We, therefore, asked whether ANO2 may be involved in this process.
GABAergic IPSCs recorded from Purkinje cells. To find out whether ANO2 is involved in this process, we compared the effect of climbing-fiber stimulation on IPSCs in wild type and ANO2 -/mice. Whole-cell recordings were obtained from sagittal tissue slices of mouse cerebellum. Purkinje cells located underneath the surface of the slice were identified by positions and shapes of their cell bodies and were selected for viability by their smooth, convex plasma membranes. A fluorescent dye was occasionally included in the pipette solution to visualize the dendritic tree (Fig 4A). Postsynaptic currents were recorded at a holding voltage V hold of -69 mV and an intracellular Clconcentration [Cl -] i of 131 mM. Detectable postsynaptic currents varied from a few pA to 819 pA (mean: 200.6 ± 5.8 pA). Their time course was characterized by a mean rise time (interval from 10% to 90% of maximal current) of 0.75 ± 0.007 ms, and a mean decay time constant (single-exponential fit) of τ = 5.54 ± 0.06 ms (Fig 4B; overlay of 764 signals). The signals were completely blocked by 50 μM picrotoxin in the bath solution ( Fig  4C) and were thus pharmacologically identified as GABAergic currents. They originated from synapses that inhibitory interneurons (basket cells, stellate cells) form on the Purkinje cell dendrites and somata [43]. The negative polarity of these signals indicates that GABA A receptors conduct Clefflux at the high intracellular chloride concentration used. In these experiments, differences in amplitude may, in part, result from different local [Cl -] i levels at each individual synapse. Moreover, electrical signals caused by dendritic postsynaptic currents in Purkinje cells are considerably attenuated and filtered as they travel along the dendrite toward the soma. With increasing distance between soma and the synaptic location on the dendritic tree, signal amplitudes decrease and time constants increase [44].  The number of detectable IPSC signals decreased by~47% through climbing-fiber stimulation (before CF: 30.7 ± 6.5 min -1 ; after CF: 14.6 ± 3.4 min -1 ; 8 cells; ctrl). In slices from Ano2 -/mice, more IPSCs were detected (54.5 ± 18.5 min -1 ; 4 cells), and the activation of climbing fibers had no effect (52.5 ± 16.2 min -1 ; 4 cells).
doi:10.1371/journal.pone.0142160.g004 ANO2 channels require more than 1 μM Ca 2+ for full activation [9,11,38]. Dendritic Ca 2+ influx was generated through synaptic activity of climbing fibers (CF) [45][46][47], which was triggered near the proximal dendrites of Purkinje cells through a stimulation pipette. CF activation designed to produce DDI (20 pulses at 1 Hz; Fig 4D) was confirmed by recording from the Purkinje cell the characteristic, prolonged Ca 2+ currents that underlie complex-spike formation [48]. [Cl -] i was set at 5 mM in these experiments (E Cl = -88 mV) to obtain inhibitory postsynaptic currents from the GABAergic synapses. At V hold between -60 mV and -45 mV, the synaptic signals were positive, indicative of postsynaptic Clinflux under this low-chloride condition (Fig 4E, upper panel). Following climbing-fiber stimulation, the recordings contained IPSC signal with reduced amplitudes (Fig 4E, lower panel). According to the concept of DDI [31], we interpret this reduced amplitude as indicative of a reduced Cldriving force for GABA-induced currents (see Discussion). To quantify the effect, we counted all IPSCs that were detected over 3 minutes. In ANO2 +/+ mice, the number of these signals decreased by 47% (8 cells) following climbing fiber stimulation (Fig 4G). The same experiments were carried out with cerebellar slices from ANO2 -/mice and revealed two differences compared to the wildtype: The number of IPSCs per minute before climbing-fiber activation was almost 2-fold higher, and it was not affected by climbing-fiber activation (Fig 4G). This observation is a strong indication for a role of ANO2 in the Ca 2+ -dependent depression of inhibition described by Satoh et al. (2013) [31].

An ANO2 channel inhibitor mimics the effect of the ANO2 knockout
A reversible ANO2 channel inhibitor would be helpful to further investigate the physiological function of ANO2 in Purkinje cells. The widely used ANO1/ANO2 blocker niflumic acid was not suitable for our experiments because it is not specific for ANO2, and it is also known to have a potentiating effect on GABA A receptors [49]. The channel inhibitor T16Ainh-A01 was originally identified as an ANO1 inhibitor in a small-molecule screen [50] and characterized in various cell types [50][51][52]. The efficacy of this compound in blocking anoctamin channels depends on the animal species and on the splice variant of the channels. In particular, T16Ainh-A01 does not effectively inhibit mouse ANO1 [52], but strongly inhibits ANO2 channels [50]. To test its suitability for our experiments on cerebellar ANO2 channels, we expressed the cerebellar variants of murine ANO1 and ANO2 in HEK293 cells and examined their sensitivity to blockage by T16Ainh-A01. Transfected HEK 293 cells were perfused with pipette solution containing 7.5 μM Ca 2+ , and the resulting currents were recorded at -70 mV immediately after whole-cell breakthrough (Fig 5A). This whole-cell method of channel blockage was used because it avoids channel alterations due to run-down, which ANO1 and ANO2 exhibit upon excision from the cell [26]. The average current densities were determined without inhibitor and compared to the values obtained in the presence of 5 μM or 25 μM T16Ainh-A01. In accordance with the report by Namkung et al. (2011) [50], the compound displayed a clear selectivity for ANO2 over ANO1 (Fig 5B). The compound did not affect currents conducted by mouse ANO1 channels at 5 μM, neither in the abc nor in the ac splice variant. Only at 25 μM T16Ainh-A01 were ANO1ac channels significantly inhibited. In contrast, ANO2-mediated currents were reduced by 27% at 5 μM and by 60% at 25 μM. As 5 μM T16Ainh-A01 selectively inhibits cerebellar ANO2 channels, we used this concentration to isolate pharmacologically any contribution of these channels to Purkinje cell activity. Although 5 μM T16Ainh-A01 exerts only a moderate inhibitory effect on ANO2, it is suitable to experimentally distinguish ANO2 effects from any contributions by ANO1. In cerebellum slices incubated in ACSF containing 5 μM of the inhibitor, we recorded significantly increased IPSC numbers per minute. This elevated IPSC frequency was resistant to stimulation of climbing fibers (Fig 5C and 5D).
In fact, the IPSC activities before and after the induction of complex spikes did not differ significantly between Ano2 -/mice and wild-type mice recorded with 5 μM channel inhibitor. These results further support the hypothesis that ANO2 mediates the plasticity of GABAergic inhibition in Purkinje cells.

Ionic plasticity may underlie the ANO2 effect on GABAergic inhibition
If ANO2 modulates IPSCs in GABAergic synapses by changing postsynaptic [Cl -] i , as suggested by the findings of Satoh et al. (2013) [31], the polarity and amplitude of IPSCs should depend on the activity of ANO2. We tested this assumption under conditions where a reversal of IPSC polarity may report an increase of [Cl -] i at synaptic sites. We used 12 mM [Cl -] i (E Cl near -62 mV) in the pipette solution and clamped the holding-voltage to -60 mV. The rationale was to record GABAergic signals with E Cl set close to V hold , so that any driving force for Clcould only result from local changes of [Cl -] i near the GABAergic synapses. Under these conditions, the postsynaptic currents had negative polarity indicating that E Cl at the synaptic sites was more positive than V hold and that GABA A receptors conducted Clefflux (Fig 6A). Current amplitudes were small as a consequence of the small driving force for chloride ions (V m-E Cl < 0). Shortly after perfusion of the ANO2 inhibitor, both positive and negative signals were recorded (Fig 6B) and, upon continuous application of the inhibitor, most postsynaptic signals were inverted to positive polarity (Fig 6C). This effect of the inhibitor was observed at various times after the start of the experiment, but the polarity switch was never observed without the inhibitor. An analysis of signal traces from 9 Purkinje cells at 12 mM [Cl -] i showed that the ANO2 inhibitor decreased the incidence of negative postsynaptic signals from 54.0 ± 11.6 min -1 to 9.0 ± 4.2 min -1 , while the incidence of positive signals increased from zero to 25.7 ± 6.0 min -1 (Fig 6D). Thus, application of the ANO2 inhibitor triggered a shift of the local E Cl from a value more positive than V hold to a value more negative than V hold , caused by a In the absence of the ANO2 inhibitor (upper scheme), the basal activity of ANO2 channels (green) provides a Clconductance in the dendritic membrane. ANO2 contributes to the Cltransport machinery, whose various pathways are represented by the K + /Clcotransporter KCC2 (blue). Together the Clpathways stabilize a slightly elevated level of [Cl -] i which results in a negative driving force (V m -E Cl < 0) for Clcurrents through GABA A receptors in GABAergic synapses (red). In this situation, Clcurrents are outwardly directed and cause negative postsynaptic currents. Application of the ANO2 inhibitor (lower scheme) reduces the Clconductance. This causes a polarity reversal of the Cldriving force, as the balance shifts towards Clextrusion, causing local [Cl -] i to decrease. This hypothesis provides a qualitative concept for the role of ANO2 channels in the inversion of postsynaptic currents that is depicted in panels A to C. The proximity of Cltransport pathways and GABAergic synapses, as well as the occurrence of local Cl -. gradients within dendritic segments, are inspired by the model for GABA A -receptor-mediated Clgradients in extended dendritic trees proposed by Jedlicka et al. (2011) [88]. It appears that, at 12 mM [Cl -] i in the pipette solution, the various chloride pathways present in the Purkinje cell dendrite uphold a chloride level that supports Clefflux through GABA A receptors (Fig 6E, upper scheme). If, however, the ANO2 Clconductance is blocked by T16Ainh-A01 over a prolonged period of time, the balance between Cluptake and Clextrusion seems to change. Local [Cl -] i decreases and promotes Clinflux through GABA A receptors (Fig 6E, lower scheme). This interpretation of the data presented in Fig 6 suggests that, even without experimentally induced climbing-fiber activation, ANO2 channels have a sufficient basal activity to influence local [Cl -] i homeostasis near the synaptic sites. This may be the consequence of some excitatory input, and hence Ca 2+ influx, in the slice preparation. These experiments suggest that ANO2 channels are a component of the transport system that regulates postsynaptic [Cl -] i in the Purkinje-cell dendritic tree. As a Ca 2+ -gated channel, ANO2 is expected to exert its effect most efficiently near sites of Ca 2+ entry. We assume that the dendritic spines are the main source of Ca 2+ entry [53], and that Ca 2+ signals are strongest in the distal dendrites. This may be due to the larger surface-to-volume ratio and a higher density of voltage-gated Ca 2+ channels in distal dendrites [54]. ANO2 in this region may contribute to DDI more efficiently than in the soma and proximal dendrite (see Discussion).
Taken together, the effects of the ANO2 inhibitor consistently demonstrate that ANO2 channels operate in the Purkinje cell plasma membrane. ANO2 appears to affect the regulation of local [Cl -] i levels in the Purkinje cell dendritic tree. Under physiological low chloride conditions, ANO2 channels mediate a Ca 2+ -dependent Cluptake into the dendrite which causes an increase of local [Cl -] i and, hence, a reduced driving force for Clentry through GABA A channels. Thus, ANO2 channels appear to mediate depolarization-induced depression of inhibitory transmission, DDI, as described by Satoh et al. (2013) [31].

ANO1 and ANO2 are expressed in the cerebellar cortex
We have examined the questions whether ANO1 and ANO2 proteins form Ca 2+ -activated Clchannels in the cerebellar cortex, and whether these channels contribute to the DDI form of synaptic plasticity. PCR experiments and immunoblots showed that both ANO1 and ANO2 are expressed. Immunosignals indicated expression of both proteins in Purkinje cells, but ANO1 also in basket, stellate and Golgi cells. We identified the splice variant ANO1ac, a variant previously shown to possess particularly high Ca 2+ sensitivity. The EC 50 for channel activation by Ca 2+ at -40 mV was reported to be 0.13 μM for ANO1ac and 0.63 μM for ANO1abc [5]. Segment a sensitizes ANO1 to Ca 2+ [55], and segment c stabilizes the open state of the channel [56]. Immunostaining of ANO1 and ANO2 in the cerebellum required the preparation of cryosections without prefixation of the brain, a protocol known to increase detection sensitivity. Total protein extracts from tissue lysates did not produce anoctamin-specific signals in immunoblots, as reported earlier for ANO2 [8]. It was necessary to remove nuclei and soluble proteins and to enrich membrane proteins. These observations point to a relatively low expression level of ANO2 in the cerebellum, consistent with our RT-PCR analysis and immunostaining. The ANO1ac protein is expressed in all GABAergic neurons of the cerebellar cortex. Its function in these cells remains to be elucidated in future studies. In Purkinje cells, ANO1 appears to be restricted to the cell body and is hardly detectable in the dendritic tree. Possibly, the channels provide a Ca 2+ -dependent component to cell volume control, as was suggested for ANO1 in epithelial cells [57]. It is, however, unlikely that ANO1ac channels contribute significantly to DDI. Because the application of 5 μM ANO2 inhibitor, as well as the ablation of the Ano2 gene, completely removed DDI, our data strongly suggest that ANO2 channels mediate DDI.
The evidence for an involvement of ANO2 in ionic plasticity presented here is based on the comparison between wild-type Ano2 +/+ mice and Ano2 -/knockout mice. The availability of the Ano2 -/mouse line [8] strengthens the validity of the data, which would otherwise have to rely on the specificity of ANO2 antisera and the ANO2 inhibitor. However, the lack of dendritic ANO2 immunosignals in Ano2 -/-Purkinje cells, together with the resistance of IPSCs to CF-stimulation in the knockout, strongly corroborate the hypothesis that ANO2 is necessary to trigger DDI. Nevertheless, it appears striking that the impact of 5 μM ANO2 inhibitor on IPSCs is strong enough to mimic the effect of an ANO2 knockout, considering that the blocker reduces ANO2 currents in HEK293 cells by only 27% at that concentration. This observation can be interpreted as a cumulative effect that developed during the extended presence of the blocker. The ANO2 inhibitor was applied for several minutes in our experiments. The continuously reduced dendritic Clpermeability during this time is expected to result in a progressive change of dendritic [Cl -] i . Thus, temporal aspects have to be considered for the evaluation of the ANO2 inhibitor data. This is illustrated by Fig 6 where the inversion of postsynaptic current polarity is seen to develop with time.
Our data indicate that ANO2 channels are active even without climbing fiber activation and despite the presence of EGTA in our pipette solution. This is surprising, considering that Purkinje cells have a particularly high Ca 2+ buffer capacity [58] and that the baseline Ca 2+ concentration is expected to be low [47,59]. It indicates that, in our preparation, local Ca 2+ fluctuations in the dendrites are sufficient to induce some degree of ANO2 activity. The isoform B of ANO2, which is expressed in Purkinje cells, is half-maximally activated at 1.33 μM Ca 2+ [6]. Thus, channel activity, detected through effects of the channel blocker in the present study, may reflect Ca 2+ concentrations of~1 μM at the expression sites of ANO2. This points to a proximity of ANO2 channels and the sites of synaptic Ca 2+ -signal generation in Purkinje-cell, the dendritic spines. It appears that the dendritic ANO2 channels respond to the Ca 2+ concentration in the spines, and that they influence the local Clconcentration inside the spines. The range of Ca 2+ -concentrations in which cerebellar ANO2 channels operate (1-10 μM) [9,11,38] corresponds well to the dynamic range of Ca 2+ transients measured in Purkinje cells. Ca 2+ concentrations of 0.5 to 4 μM induce long-term depression upon activation of climbing fibers [59]. These matching Ca 2+ ranges suggest that the open probability of dendritic ANO2 channels increases when excitatory synaptic inputs generate dendritic calcium signals through activation of voltage-gated Ca 2+ channels and Ca 2+ release (reviewed by [54]). ANO2 appears to be a Ca 2+ -dependent element of chloride homeostasis in Purkinje cell dendritic spines. In concert with the chloride exporter KCC2 [60,61], the channels appear to set local levels of [Cl -] i and, hence, to co-determine the efficacy of GABAergic transmission. A contribution of the chloride importer NKCC1 was suggested [31] because 50 μM bumetanide reduced an initial, transient phase of DDI that was observed in that study. However, expression of NKCC1 in Purkinje cells is not established; several studies report the absence of NKCC1 mRNA from Purkinje cells in the adult cerebellum [62][63][64].

ANO2 channels and synaptic plasticity in the cerebellum
The formation of procedural memory, in particular the acquisition of motor skills, involves the cerebellum as central element in the preparation, the execution, the fine adjustment, and in the learning of movement. The characteristically uniform microcircuits in the cerebellar cortex and the deep nuclei constitute the hub of signal processing for these tasks [65][66][67][68]. It is well documented that plasticity of inhibition in GABAergic synapses is a major factor in shaping the output pattern of the cerebellar cortex [69][70][71]. The sites of plasticity are the synapses formed by inhibitory interneurons on the Purkinje cell dendrites and somata. Various distinct modes of synaptic plasticity control the efficacy of inhibitory transmission in these synapses: rebound potentiation (RP), a postsynaptic Ca 2+ / CaMKII-mediated process that transiently enhances the responsiveness of GABA A receptors upon depolarization [72][73][74][75]; depolarisationinduced potentiation of inhibition (DPI), a delayed, long-lasting increase of GABAergic transmission mediated by presynaptic NMDA-receptors [76]; depolarization-induced suppression of inhibition (DSI), a reduction of IPSPs for~20 s induced by retrograde inhibition of GABAergic synapses through endocannabinoids [37,[77][78][79]; depolarization-induced depression of inhibition (DDI), a decrease of IPSC amplitudes as a consequence of increasing postsynaptic Clconcentration, mediated by dendritic Ca 2+ -activated Clchannels and, possibly, Cltransporters [31]. Taken together, these modes of synaptic plasticity reflect the relevance of modulatory processes for GABAergic inhibition of Purkinje cell activity.
The Ca 2+ -dependent Clconductance observed in Purkinje cells [77] was previously shown to be involved in the generation of DDI [31]. DDI changed the amplitudes of IPSCs that were evoked by stimulation of interneurons in the molecular layer and mediated by GABAergic synapses. The basic observation was that the postsynaptic currents transiently changed polarity from outward to inward (i.e. from inhibitory to excitatory), when a series of depolarizing pulses was applied to the Purkinje cells. Even after partial recovery and restoration of outward polarity, the IPSC amplitudes remained small for over 20 min. Satoh et al. (2013) [31] presented evidence that DDI was caused by a Ca 2+ -dependent rise of [Cl -] i on the postsynaptic side (in the Purkinje cell dendrite), resulting in a decreased driving force for Clinflux through GABA A receptors. As such, DDI would be an example of ionic plasticity [80][81][82][83], an altered synaptic transmission, brought about not by GABA A receptor regulation but because of a changing driving force for Clflux.
In the present study, changes in [Cl -] i were inferred from the effect of the ANO2 blocker on the amplitude of postsynaptic currents in GABAergic synapses. The interpretation of this indirect read-out is based on the assumption that the amplitudes of GABA-induced postsynaptic currents depend on [Cl -] i and mirror any change of [Cl -] i . A decline of IPSC amplitude may, however, also be caused by changes in the neurotransmission process or by changes in the passive electrical properties of the dendrite. Such an electrical effect was proposed for ANO2 activation near excitatory synapses in hippocampal pyramidal neurons [27]. However, changes of GABA transmission in Purkinje cells following CF-activation occur on much slower time scales than used in this study [84] and require phosphorylation and recruitment of new receptor subunits. Moreover, it is unlikely that the decrease of IPSC amplitudes after CF-activation results from passive electrical effects like the introduction of a shunt conductance. This can be inferred from the effects of CF-activation on EPSPs at parallel-fiber synapses on Purkinje-cell dendrites. Time course and amplitude of EPSPS are not altered by CF-activation, provided that retrograde inhibition is blocked by a cannabinoid-receptor antagonist [85]. Thus, within the context of DDI, it is reasonable to assume that decrease and inversion of IPSCs observed here result from changes in [Cl -] i . Nevertheless, it would be desirable to directly measure changes of [Cl -] i by chloride imaging, a method that was successfully applied to Purkinje cell dendrites using two photon excitation of the fluorescent dye MQAE [86]. This is technically demanding but may be a suitable approach to assess the contribution of ANO2 channels to [Cl -] i regulation quantitatively.
The neuronal chloride homeostasis is mediated to a large extent by cation-chloride cotransporters, in particular by KCC2 and, in some neurons, by NKCC1 [87]. Chloride channels can, however, transiently change [Cl -] i , especially within the restricted space of distal dendrites [88][89][90]. This was studied in detail for the dendrites of hippocampal pyramidal neurons, where, as a consequence of intense of GABA A receptor activity, [Cl -] i increases and depolarizes E Cl [91][92][93]. This GABA-induced Cluptake can overcome the intrinsic Clextrusion mechanism in parts of the dendrite with high density of inhibitory synapses. It can change E Cl locally, and can hence cause a spatially restricted attenuation-or even inversion-of GABA effects [81]. In a similar way, ANO2 channels may affect [Cl -] i and attenuate GABAergic input in the Purkinje cell dendrite, when activated by Ca 2+ near excitatory synapses. In contrast to the homeostatic regulation of [Cl -] i by cation-chloride cotransporters, the effect of ion channels on dendritic E Cl ceases after Clchannels close and Clextrusion prevails. Satoh et al. [31] reported, that Purkinje cells largely recovered from DDI within 2-3 minutes, but that inhibition was still slightly attenuated 20 min after DDI induction. If ANO2 channels mediate DDI, this slow time course of recovery reflects three processes following the end of CF-activation: return of dendritic Ca 2+ levels to sub-micromolar levels, closing of ANO2 channels, and Clextrusion from the dendrite. Ca 2+ -imaging studies during CF-activation indicated that the Ca 2+ decline proceeds within less than a second [58], and the open probability of ANO2 follows changes in Ca 2+ concentration within a fraction of a second [11]. This suggests that the recovery dynamics of DDI are determined by the net rate of Clextrusion. This rate may be determined by Climaging in future studies.
In conclusion, we have analyzed the expression of ANO1 and ANO2 Ca 2+ -activated Clchannels in the cerebellar cortex. While ANO1 is expressed in inhibitory interneurons and in the somata of Purkinje cells, ANO2 is absent from inhibitory interneurons but specifically targeted to the Purkinje cell dendritic tree. ANO2 channels have previously been studied in dendritic signal processing of olfactory receptor neurons [6][7][8][9][10] and hippocampal pyramidal neurons [27] as well as in the pre-synaptic terminals of rod photoreceptors [14,15,94]. Here we provide evidence that dendritic ANO2 channels co-determine local chloride concentrations near GABAergic synapses. ANO2 channels appear to mediate DDI, a Ca 2+ -dependent form of ionic plasticity that attenuates GABAergic inhibition in cerebellar Purkinje cells.