Anoctamin 2-chloride channels reduce simple spike activity and mediate inhibition at elevated calcium concentration in cerebellar Purkinje cells

Modulation of neuronal excitability is a prominent way of shaping the activity of neuronal networks. Recent studies highlight the role of calcium-activated chloride currents in this context, as they can both increase or decrease excitability. The calcium-activated chloride channel Anoctamin 2 (ANO2 alias TMEM16B) has been described in several regions of the mouse brain, including the olivo-cerebellar system. In inferior olivary neurons, ANO2 was proposed to increase excitability by facilitating the generation of high-threshold calcium spikes. An expression of ANO2 in cerebellar Purkinje cells was suggested, but its role in these neurons remains unclear. In the present study, we confirmed the expression of Ano2 mRNA in Purkinje cells and performed electrophysiological recordings to examine the influence of ANO2-chloride channels on the excitability of Purkinje cells by comparing wildtype mice to mice lacking ANO2. Recordings were performed in acute cerebellar slices of adult mice, which provided the possibility to study the role of ANO2 within the cerebellar cortex. Purkinje cells were uncoupled from climbing fiber input to assess specifically the effect of ANO2 channels on Purkinje cell activity. We identified an attenuating effect of ANO2-mediated chloride currents on the instantaneous simple spike activity both during strong current injections and during current injections close to the simple spike threshold. Moreover, we report a reduction of inhibitory currents from GABAergic interneurons upon depolarization, lasting for several seconds. Together with the role of ANO2-chloride channels in inferior olivary neurons, our data extend the evidence for a role of chloride-dependent modulation in the olivo-cerebellar system that might be important for proper cerebellum-dependent motor coordination and learning.


Introduction
The activity of a neuronal network depends on the excitability of the neurons within this network and on the modulation of this excitability. Excitability of individual neurons can either be increased or decreased, and both changes will affect the output signal of the network. Calcium-activated chloride channels are of special interest in this context, because they can either enhance or attenuate neuronal excitability in various ways (see [1] for recent review). The intracellular chloride concentration differs between neurons and can vary between different compartments of the same neuron (reviewed in [2]). Consequently, calcium-induced chloride currents can be either hyperpolarizing or depolarizing. This characteristic feature of chloride currents leads to various modes of modulation through which excitability can be affected. The calcium-activated chloride channel Anoctamin 2 (ANO2) was found to decrease excitability in neurons of the hippocampus [3], thalamus [4], lateral septum [5], and central lateral amygdala [6], whereas excitability is increased in olfactory receptor neurons [7,8] and the olivocerebellar system [9,10]. The enhancing effects of ANO2 in the olivo-cerebellar system has been associated with impairments in motor coordination and learning tasks observed in ANO2-deficient mice (Ano2 -/-) [9,11]. However, while the enhancement of climbing fiber signals by ANO2 is well established, the effect of ANO2 on Purkinje cell excitability remains controversial.
In climbing fibers, ANO2 facilitates rapid repolarization after high-threshold spikes and, therefore, enables faster spiking [9]. Additionally, ANO2 has been suggested to attenuate GABAergic inhibition from molecular layer interneurons onto Purkinje cells [10]. Because climbing fibers are also involved in this process [12], it is still unclear whether ANO2 in Purkinje cells or in climbing fibers is responsible for the reduction of GABAergic signals. In the present study, we specifically examine the contribution of ANO2 to the modulation of excitability in Purkinje cells. To this end, we analyze the excitability of Purkinje cells in acute tissue slices, uncoupled from climbing fiber input, and compare wildtype to Ano2 -/mice. Although we used a global Ano2 -/mouse for this study, our experimental design revealed local effects of ANO2 on Purkinje cell neurons and illustrated the impact of ANO2 solely on the network of the cerebellar cortex.
Our results confirm, that ANO2 is expressed in Purkinje cells. We discovered that ANO2 decreases instantaneous simple spike activity of Purkinje cells following strong current injections and current injections close to the activation threshold of Purkinje cells. Moreover, we report an ANO2-mediated attenuation of GABAergic inhibition for a few seconds after depolarization of Purkinje cells. Combined with the proposed role of ANO2 in climbing fibers [9], we integrate these new findings into the concept of chloride-dependent modulation of network activity in the cerebellar cortex.
Scientific, Waltham, Massachusetts, USA) or the RNeasy Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. cDNA was synthetized with 50 ng of each RNA preparation using the SuperScript™ VILO™ Master Mix (Thermo Fisher Scientific) according to the manufacturer's instructions. Quantitative real-time PCR (qPCR) was performed with StepOne Real Time PCR System (Applied Biosystems, Foster City, California, USA) using the Fast Start Universal SYBR Green Master (ROX) (Roche, Basel, Switzerland). In every run, a negative control and duplicates for each primer pair were used. A melting-curve analysis was included in every qPCR run. YWHAZ and Gusb were chosen as housekeeping genes, as they are supposed to be most consistently expressed across different regions of the mouse brain [13].
The CT-values, obtained from the StepOne Real Time PCR System, were analyzed using the relative quantification method [14] which is implemented in the REST 2009 software (Qiagen). Data obtained from inferior olivary tissue samples were chosen as reference sample. Primer efficiencies were calculated using the Miner-Algorithm [15]. Since the resulting foldexpression values are not normally distributed, the 68% confidence was calculated in REST 2009 using a randomization approach and was used as a measure of variation of the samples.

In situ hybridization (ISH)
cDNA was synthesized from extracted RNA from olfactory epithelium or cerebellum using the SuperScript™ VILO™ Master Mix (Thermo Fisher Scientific). cDNA fragments were amplified through PCR using two sets of primers (ano2_1: primer pair used for qPCR (see above), ano2_2/ f ACAACGGGACACTACATGGC, ano2_2/r TCGCTGATGTCAGTGGGGAT). The PCR products were purified using the PCR Cleanup Kit (Macherey-Nagel, Düren, Germany) and ligated in a pGEM-T Vector (Promega, Madison, Wisconsin, USA). Once ligated, plasmids were transformed through heat-shock into JM109 High-Efficiency E. Coli Competent Cells (Promega). Positive recombinants were incubated in LB-Medium and the samples were sequenced in Microsynth SeqLab (Göttingen, Germany) after minipreparation. After confirming the proper ligation of the insert and match with Ano2 mRNA through BLAST, the plasmids were linearized using the enzymes NcoI or NdeI (Thermo Scientific). The linearized vectors were finally incubated for 2 h at 37˚C with the RNA-Polymerases SP6 or T7 using the DIG-RNA-Labeling Mix (Roche) to synthesize the sense (control) and the antisense hybridization probes, respectively. C57BL/6N mice were transcardially perfused with 4% paraformaldehyde (PFA). After perfusion, heads were fixed for 2 h in 4% PFA and then washed with PBS (150 mM NaCl, 8.1 mM Na 2 HPO 4 x 2H 2 0, 1.9 mM NaH 2 PO 4 x H 2 0, pH adjusted to 7.4) for 5 min and submerged in 0.5 M EDTA for at least a week at 4˚C for decalcification, until cranial bones softened. To dewater the tissue, heads were put in 10% sucrose solution (2-3 h, RT), 30% sucrose solution (overnight, 4˚C) and in a 1:1 mixture of Tissue-Tek1 O.C.T.™ Compound (2-3 h, RT) (Sakura Finetek, Tokyo, Japan) and 30% sucrose solution (2-3 h, RT), before being embedded in Tissue-Tek1 O.C.T.™ Compound only. Sagittal and coronal sections (16 μm) were sliced with a cryostat (CM3050, Leica, Wetzlar, Germany), set at a chamber temperature of -22˚C and an object temperature of -20˚C. The sections were collected on SuperFrost Plus adhesion slides (Thermo Fisher Scientific) and were left to dry for no longer than 30 min and stored at -20˚C.
As a pretreatment of the tissue, slices were fixed in 4% PFA for 10 min and then washed 2x 5 min in PBS. Afterwards, slices were digested with 50 μg/ml Proteinase K in PK-Buffer (50 mM TRIS/HCl and 5 mM EDTA adjusted to pH 8.3) for 5 min at 37˚C and washed twice in PBS for 5 min. The samples were acetylated with 0.1 M triethanolamine solution containing 0.25% acetic acid (pH 8) for 10 min. After two washing steps with PBS, slices were incubated for 4-6 h in a prehybridization buffer (25% 5x SSC, 10% 1x Denhardt's solution, 50% deionized formamide, 0.1% Tween-20, 500 μg/ml Herring Sperm DNA; for 20xSSC: 3 M NaCl, 0.3 M trisodium citrate, pH 7.0) before being incubated overnight with the hybridization buffer (25% 5x SSC, 10% 1x Denhardt's solution, 50% deionized formamide, 0.1% Tween-20) containing 350-450 ng of the hybridization probe at 55˚C. On the next day, the slices were washed with 5x SSC at 55˚C, 0.2x SSC at 55˚C and 0.2x SSC at RT for 30 min each. Slices were washed 5 min in PBS with 0.1% Triton X-100 and 0.5% BSA (PBT) and then blocked with a blocking solution containing 5% BSA in PBT for 1 h at RT. After blocking, the anti-Digoxigenin-AP antibodies (Roche, dilution 1:5000 in blocking solution) were added and incubated for 2 h at RT. Slices were washed 2x 30 min in PBT and incubated in NBT/BCIP coloration solution (Ready-to-use tablets, Roche) until a staining was visible for the antisense but not the sense hybridization probe (30 min to 4 h or overnight). Slices were washed with PBS to stop the reaction. The inferior olivary nucleus (coronal slices) or the inferior olivary nucleus and the hippocampus (sagittal slices) served as positive controls in each staining. Every staining procedure included stainings with the sense hybridization probe as negative control.
Images of ISH stainings were taken with an Eclipse 90i microscope (Nikon, Minato, Japan) using the brightfield camera DS-Ri1 (Nikon) and IHC stainings were captured with the confocal laser scanning system D-Eclipse C1 (Nikon). Images were captured with the imaging software NIS-Elements AR 4.00.12 (RRID:SCR_014329, Nikon). The following objectives were used: 4x (Nikon Plan Fluor 4x/NA 0.13) and 60x (Nikon Plan Apo VC 60x/NA 1.40 Oil). Contrast and brightness of the images were adjusted with CorelPHOTO-PAINT 2017 (Corel Corporation, Ottawa, Canada).

Free-floating immunohistochemistry
The cerebellum was removed and fixed for 30 min in 4% paraformaldehyde (PFA) solution. The cerebellum was cut in ice-cold PBS into 200 μm-thick slices with a vibratome (VT1000S, Leica). The slices were fixed again for 30 min in 4% PFA solution. After washing for 4x 10 min with PBS, slices were incubated for 2 h in a blocking solution containing 5% goat serum (Sigma-Aldrich), 0.5% Triton X-100 and 0.05% NaN 3 . Primary antibodies were applied overnight (diluted in blocking solution) and sections were washed again 4x 10 min in PBS with 0.5% Triton X-100 (PBST). Unless the primary antibody was directly coupled with a fluorescent tag, the slices were incubated for 2 h with the secondary antibodies (diluted in blocking solution), which were conjugated to a fluorescent tag. After washing 4x 10 min with PBST, slices were mounted on glass slides with Aqua-Poly/Mount (Polysciences). The slices were kept on a shaker at RT in the dark for the entire procedure.
Images of IHC stainings were taken with an Eclipse 90i microscope (Nikon) using the confocal laser scanning system D-Eclipse C1 (Nikon). Images were captured with the imaging software NIS-Elements AR 4.00.12 (Nikon) using a 40x (Nikon Plan Fluor 40x/NA 1.30 Oil) and a 60x (Nikon Plan Apo VC 60x/NA 1.40 Oil) objective. For analysis of Purkinje cell morphology, density of Purkinje cells and molecular layer interneurons, z-stacks (range 4-9 μm) were taken with a 40x objective and the maximum intensity projections were analyzed with ImageJ (RRID:SCR_003070, NIH, Bethesda, Maryland, USA). Interneurons were identified by subtracting the calbindin signal from the parvalbumin signal [25]. Because the cell bodies of Purkinje cells are aligned in a row, density of Purkinje cells was defined as cell bodies per mm. On the other hand, density of molecular layer interneurons was defined as cells per volume of the molecular layer, since they are scattered throughout the molecular layer. For counting climbing fiber and GABAergic synapses, images were pre-processed with iLastik (RRID: SCR_015246, HCI, Heidelberg, Germany) and then further analyzed with ImageJ. Images were taken at 60x magnification and a zoom factor ranging from 1.0 to 2.0. As climbing fiber synapses are located primarily on the main dendrites, only synapses in the proximal half of the molecular layer were counted. To analyze the distribution of climbing fiber synapses, distances between adjacent synapses on or next to the same dendrite were measured. Data are presented as curves describing the shape of the underlying histogram (bin size 0.5 μm). Contrast and brightness of the images shown in this study were adjusted with CorelPHOTO-PAINT 2017 (Corel Corporation).

Electrophysiology
The cerebellum was removed immediately after sacrificing the animal and was mounted on the stage of a vibratome (VT1000S, Leica) with cyanoacrylate. The tissue was covered with artificial cerebrospinal fluid (ACSF) containing 134.6 mM NaCl, 21 mM NaHCO 3 , 3.4 mM KCl, 0.6 mM NaH 2 PO 4 , 1 mM MgCl 2 , 2.5 mM CaCl 2 . The pH was adjusted to 7.4 with HCl and osmolarity was set between 305 and 320 mOsmol/kg. During slicing, the temperature of the ACSF was set to 32-35˚C. Acute parasagittal cerebellar slices (200-250 μm) were obtained and incubated in a holding chamber at 35˚C for at least 30 min, as previously reported [26,27]. During the experiments, the holding chamber was kept at RT (25-28˚C). All solutions were constantly oxygenated with 95% O 2 /5% CO 2 .
For electrophysiological experiments, the slices were transferred onto the stage of an upright microscope (Nikon Eclipse E600FN). The recording chamber was continuously perfused with oxygenated ACSF at a speed of 1-2 ml/min at RT. Purkinje cells were identified by their morphology using a brightfield camera (DS-Ri1, Nikon) and a water-immersion objective (Nikon Fluor 40x/ 0.80W). For all experiments, borosilicate glass micropipettes (Science Products, Hofheim, Germany, GB-150F 10) were produced with a horizontal puller (P97; Sutter Instruments, Novato, California, USA) and polished with a microforge to a resistance of 3-5 MO. Electrical signals were recorded using a patch-clamp amplifier (EPC 8, HEKA Electronics, Lambrecht, Germany) and the WinWCP Software (RRID:SCR_014713, Version 5.5.2, University of Strathclyde, Glasgow, UK).
For current-clamp experiments, micropipettes were filled with a K-methane sulfonate based internal solution containing 150 mM K-methane sulfonate, 5 mM KCl, 0.1 mM Cs-EGTA, 10 mM Na-HEPES, 3 mM MgATP, 0.4 mM NaGTP, pH 7.4 with KOH, osmolarity 302 mOsmol/kg. This resulted in a reversal potential for chloride (E Cl ) of -87 mV that has previously been reported to reflect the physiological conditions of Purkinje cells [28]. Mature Purkinje cells have been shown to express KCC2 but not its counterpart NKCC1 [29,30], leading to a pronounced chloride extrusion and, therefore, a low E Cl .
After obtaining the whole-cell configuration, the cell was held at -70 mV for 5 min. Two test pulses (-5 mV and -10 mV, 0.5 s) were induced in voltage-clamp mode to analyze series and input resistance off-line. Spontaneous activity was recorded in current-clamp mode (0 pA, 1 min). To examine the intrinsic excitability of Purkinje cells, the cells were held near -70 mV with hyperpolarizing current injections that were on average comparable between wildtype and Ano2 -/mice (wt: -496.0 ± 123.5 pA; Ano2 -/-: -506.4 ± 105.8 pA). Two hyperpolarizing (-200 pA and -100 pA) and five depolarizing current pulses (200-1000 pA) were applied for 5 s to monitor the excitability of Purkinje cells in dependence of the amplitude of the current injection. For analysis, three different stages of activation were defined: threshold activation (current injection of 200 pA), moderate activation (600 pA) and strong activation (1000 pA). Each current injection was followed by a test pulse (-50 pA, 1 s) to monitor input resistance during the recording. Current pulses were separated by a break of 2 min. Data were filtered at 3 kHz and sampled at 20 kHz. Series resistance was 5-10 MO and was not compensated. To examine the role of chloride currents for excitability of Purkinje cells, measurements were repeated with a slightly modified intracellular solution, where the chloride concentration was adjusted to 23 mM chloride (132 mM K-methane sulfonate, 23 mM KCl, 0.1 mM Cs-EGTA, 10 mM Na-HEPES, 3 mM MgATP, 0.4 mM NaGTP, pH 7.4 with KOH, osmolarity 304 mOsmol/kg) to raise E Cl to -48 mV which corresponds to the mean simple spike threshold (S2D For voltage-clamp experiments, micropipettes were filled with an internal solution where K-methane sulfonate was exchanged by Cs-methane sulfonate (150 mM Cs-methane sulfonate, 5 mM KCl, 0.1 mM Cs-EGTA, 10 mM Na-HEPES, 3 mM MgATP, 0.4 mM NaGTP, pH 7.4 with CsOH, osmolarity 295 mOsmol/kg). Since hydrogencarbonate permeability has been reported for Anoctamin-chloride channels [31,32], 0.5 mM of the pH-sensitive dye HPTS (Sigma-Aldrich Cat# H1529; [33]) was added to the intracellular solution in all experiments to visualize Purkinje cell dendrites. Additionally, HPTS fluorescence was used in a preliminary experiment to confirm that ANO2 conducts mainly chloride instead of hydrogencarbonate [31,32] in our experiments. After obtaining the whole-cell configuration, cells were held at -70 mV for 10 min to ensure uniform distribution of HPTS in the dendritic tree. Inhibitory postsynaptic currents (IPSCs) were induced using a slightly modified version of the protocol from [12]. A stimulation pipette (2-2.5 MO) filled with ACSF was placed in the molecular layer. The HPTS-fluorescence signal was used to detect non-fluorescent areas between the dendrites of the Purkinje cell of interest, which were used to predict the positions of molecular layer interneurons. Electrical stimulation using a constant current source (DS7A; Digitimer, Welwyn Garden, UK) induced evoked IPSCs (eIPSCs) every 15 s (800 μA, 100 μs, interstimulus interval 50 ms) at a holding potential of -60.5 mV. Experiments on the reduction of inhibition were performed, as previously described [12], with a train of five depolarizing pulses via the patch pipette (-70.5 mV to -20.5 mV for 1 s at 0.5 Hz). A test pulse (5 mV, 0.5 ms) was induced before each stimulation to monitor the series resistance, which was 5-9 MO in all experiments. Series resistance was not compensated. Data were filtered and sampled at 3 kHz. ACSF included the AMPA/Kainate receptor antagonist CNQX (10 μM, Tocris Bioscience, Bristol, UK), the NMDA antagonist D-AP5 (40 μM, Tocris Bioscience), the GABA B antagonist CGP 55845 (2 μM, Tocris Bioscience) and the CB 1 antagonist AM 251 (2 μM, Tocris Bioscience) in all recordings. For control experiments, tetrodotoxin (TTX; 0.5 μM, Abcam) or gabazine (20 μM, Tocris Bioscience) were applied by bath application.
Electrophysiological data were analyzed off-line using Origin (RRID:SCR_014212, Version 9.0, OriginLab, Northampton, Massachusetts, USA) and Spike2 (CED, RRID:SCR_000903, Cambridge, UK). In all experiments, measurements were discarded if the series resistance changed by more than 20% during the recording. Current clamp measurements of spontaneous simple spike activity were only included in analysis, if continuous simple spike firing could be observed for 1 min. The first eight seconds were used for calculation of simple spike frequency in all recordings. Current clamp measurements with current injections were only used for analysis if the input resistance varied by less than 25% during recording and the holding potential stayed between -65 and -75 mV. Simple spikes were counted in the first 500 ms of each current injection step. For the analysis of spike frequency adaptation, interspike intervals (ISIs) of the first sixteen simple spikes after current injections were examined. The amplitude of the hyperpolarization after the current injection was calculated as the difference between the minimum membrane potential during hyperpolarization and the resting membrane potential before the current injection.
The eIPSC amplitudes were calculated as difference between the maximal current during the eIPSC and the holding current immediately before the stimulation. To compare eIPSC measurements of different cells, eIPSC amplitudes were normalized to the average eIPSC amplitude before the depolarization pulse (baseline). Time constant τ for recovery of eIPSC amplitude after depolarization was obtained by using the following model for exponential regression:

Statistics
Statistics were performed using Origin and SigmaPlot (RRID:SCR_003210, Version 13, Systat Software, San José, California, USA). All data sets were tested for normal distribution (Shapiro-Wilk test) prior to statistical analysis. Not normally distributed data were analyzed with the nonparametric Wilcoxon-Mann-Whitney test or the Kruskal-Wallis test. Normally distributed data were analyzed using Student's unpaired two-tailed t-test with Welch correction or one-way ANOVA with Holm-Sidak's multiple-comparison test, unless otherwise stated. Twoway ANOVA with Holm-Sidak's multiple-comparison test was used for the comparison of interspike intervals. Significance level was set to 0.05. Sample sizes were defined as n = number of images or cells from N animals. Data are presented as box plots, with the line indicating the median, the square indicating the mean and the whiskers representing 1.5x interquartile range (IQR). In the figure legends, normally distributed data are presented as mean ± SD otherwise as median with IQR.

Ano2 is expressed in cerebellar Purkinje cells
An expression of ANO2 in Purkinje cells has previously been reported using immunohistochemistry [10,11]. On the other hand, no fluorescence was detected in Purkinje cells of another global ANO2-knockout mouse model that expresses farnesylated mCherry instead of ANO2 [9]. This raised a controversy about whether ANO2 is indeed expressed in Purkinje cells and, to examine this question, we performed qPCR and in situ hybridization experiments to investigate the expression of Ano2-mRNA in cerebellar tissue. Double-stainings of in situ hybridization and immunohistochemistry were used to verify the identity of the neurons expressing Ano2.
Our qPCR experiments revealed an expression of Ano2 mRNA in both cerebellar hemispheres and the Vermis (S1 Fig). In situ hybridization was performed with two different hybridization probes for Ano2 in independent experiments. Both hybridization probes provided a robust signal along the Purkinje cell layer (PCL) in all parts of the cerebellum (Fig 1A). No signal was detectable in the control stainings with the two sense probes (Fig 1A inset). The signals for in situ hybridization (Fig 1B, upper) and immunohistochemistry for calbindin, a marker for Purkinje cells (Fig 1B, middle), clearly overlapped (Fig 1B, lower). These data demonstrate that Ano2 is expressed in cerebellar Purkinje cells.

Ano2 -/mice
Most experiments in this study are based on the comparison of wildtype mice and mice lacking ANO2-chloride channels (Ano2 -/mice). To check for neuroanatomical integrity of the cerebellar cortex in Ano2 -/mice, we analyzed density and size of Purkinje cells, as well as the density of molecular layer interneurons with double immunohistochemical stainings of the Purkinje cell marker calbindin together with parvalbumin that stains both Purkinje cells and molecular layer interneurons (Fig 2A). Additionally, we investigated the density of interneuron synapses on the Purkinje cell dendrites, as well as the density and distribution of climbing fiber synapses with co-stainings for calbindin, VGAT and VGluT2, to visualize Purkinje cells, GABAergic and climbing fiber synapses in the molecular layer of wildtype and Ano2 -/mice, respectively (Fig 2E and 2F).
We did not observe any significant differences between wildtype and Ano2 -/mice, neither for density (Fig 2B) or size of Purkinje cells (Fig 2C), nor for density of molecular layer interneurons ( Fig 2D). As anticipated, GABAergic synapses covered the entire molecular layer, whereas climbing fiber synapses only populated the proximal two thirds of the molecular layer in both wildtype and Ano2 -/mice. We did not detect any significant difference between wildtype and Ano2 -/mice for the density of GABAergic synapses (Fig 2G) nor for the density of climbing fiber synapses ( Fig 2H). Climbing fiber synapses were regularly distributed in both wildtype and Ano2 -/mice, with distances typically ranging from 1 to 5 μm, without clusters of synapses being observable (Fig 2I).
These results provide evidence, that the neuroanatomical parameters of the cerebellar cortex of Ano2 -/mice are not significantly altered compared to wildtype mice. Therefore, we can assume that functional differences between wildtype and Ano2 -/mice are due to the absence of ANO2 chloride-channels as opposed to being a consequence of neuroanatomical changes in the cerebellar cortex of Ano2 -/mice.

ANO2 does not affect spontaneous simple spike activity of Purkinje cells
Purkinje cells display intrinsically driven simple spike discharge at high rates (ca. 40-50 Hz) which is modulated by parallel fiber input [34,35]. Additionally, Purkinje cells receive excitatory input from climbing fibers that irregularly elicit complex spikes in Purkinje cells at low rates (ca. 1 Hz) [36]. We performed current-clamp measurements on Purkinje cells in acute cerebellar slices to analyze spontaneous simple spike discharge of Purkinje cells in wildtype and Ano2 -/mice. To monitor this spontaneous activity, Purkinje cells were held at 0 pA.
The measured spontaneous simple spike firing rates of wildtype mice without gabazine corresponded to published data [35] and no significant difference was observed between wildtype and Ano2 -/mice (Fig 3A). We added 20 μM gabazine to the bath solution to eliminate

PLOS ONE
chloride currents conducted by GABA A -mediated channels. As expected, spontaneous simple spike activity became more regular and firing rates increased but were still comparable between wildtype and Ano2 -/mice (Fig 3B). These results suggest, that ANO2 does not modulate spontaneous simple spike discharge of Purkinje cells.

ANO2-mediated chloride currents reduce excitability of Purkinje cells
Spontaneous simple spike activity of Purkinje cells can be modulated by inhibitory and excitatory input from molecular layer interneurons and parallel fibers, respectively, leading to simple spike rates of up to 250 Hz [37,38]. To investigate the excitability of Purkinje cells, we first applied hyperpolarizing currents to clamp Purkinje cells close to -70 mV and then injected depolarizing current pulses of 200 to 1000 pA amplitude. Current injection of 200 pA resulted in threshold activation [39], 600 pA in moderate activation and 1000 pA in strong activation of Purkinje cells. Simple spikes were counted in the first 500 ms of each current pulse. Additionally, spike frequency adaptation as well as the hyperpolarization following the current injections were analyzed. Because ANO2-mediated changes in excitability could either be caused by chloride currents or, alternatively, by a shunt conductance via ANO2, the recordings were repeated with a slightly modified intracellular solution, where the chloride concentration was adjusted to raise the reversal potential for chloride currents (E Cl ) from -87 mV to -48 mV. As -48 mV corresponded to the mean simple spike threshold (S2D Fig), this approach resulted in a minimized net chloride current at the simple spike threshold, whereas the shunt conductance was still effective. Therefore, this second set of experiments allowed to distinguish between effects that were caused by chloride currents via ANO2 or by an ANO2-mediated shunt conductance. All recordings were performed in the presence of 20 μM gabazine, to eliminate GABA A -mediated chloride currents.
As expected, threshold activation of Purkinje cells of wildtype mice induced only a few simple spikes or no spikes at all (Fig 4A left). In contrast, repetitive simple spike activity could regularly be observed in Purkinje cells of Ano2 -/mice upon the same current injection intensity. This resulted in a significantly increased average number of simple spikes upon threshold activation in Ano2 -/mice (Fig 4A middle). Repetitive simple spike activity was also significantly

PLOS ONE
increased in Ano2 -/mice upon strong current injections (Fig 4C), whereas only a slight tendency towards increased simple spike activity could be observed in Ano2 -/mice upon moderate current injections (Fig 4B). Interestingly, these differences between wildtype and Ano2 -/mice disappeared upon elevation of E Cl to the simple spike threshold. Moreover, the elevation of E Cl appeared to only affect the simple spike activity of wildtype but not Ano2 -/-Purkinje cells. The elevation of E Cl neither influenced the basic electrical properties of Purkinje cells, Spike frequency adaptation was examined by analyzing the progression of interspike intervals (ISIs) for moderate and strong current injections. Because threshold activation did often induce only few simple spikes, spike frequency adaptation was not analyzed in this case. At physiological E Cl , the first ISIs were similar in wildtype and Ano2 -/mice for both moderate and strong activation but then gradually diverged, with ISIs growing larger in wildtype compared to Ano2 -/-Purkinje cells (Fig 5 left). In contrast, when the driving force for chloride was reduced no divergence of ISIs was observed for both moderate and strong current injections (Fig 5 right). Only a minimal parallel shift between wildtype and Ano2 -/mice were observed upon strong activation of Purkinje cells. Furthermore, we analyzed the extent and time course of hyperpolarization that occurred after returning to the holding current that clamped Purkinje cells close to -70 mV. The amplitude of afterhyperpolarization was significantly decreased in Ano2 -/mice compared to wildtype following strong activation at physiological, but not elevated E Cl (Fig 6B). The same tendency, although not statistically significant, was observed after moderate and threshold activation (S3 Fig). No difference between wildtype and Ano2 -/mice was observed for the time until maximal afterhyperpolarization, neither at physiological nor at increased intracellular chloride concentration.
Taken together, these results indicate that ANO2-mediated chloride currents reduce the excitability of Purkinje cells at the activation threshold and during strong activation. These chloride currents lead to a prolongation of ISIs from the second simple spike on and additionally promote hyperpolarization after current injections stop.

ANO2 attenuates inhibition after depolarization of Purkinje cells
It has previously been suggested that ANO2 is involved in a process causing attenuation of GABAergic inhibition following climbing fiber input [10]. This process was proposed to rely on a calcium-mediated influx of chloride into the dendrites that increases the local intracellular chloride concentration and, thus, diminishes the driving force for chloride across the dendritic membrane [12]. The resulting reduction of amplitude observed in spontaneous inhibitory postsynaptic currents (IPSCs) at GABAergic synapses of wildtype mice, could not be detected in Ano2 -/mice [10]. This finding gave rise to the idea that ANO2 would temporarily relieve Purkinje cells from GABAergic inhibition after a calcium surge that accompanies a complex spike. However, it is still unclear whether this effect is based on ANO2 channels in inferior olivary neurons or ANO2 channels in the dendritic membrane of Purkinje cells.
To address this question, we performed whole-cell voltage clamp recordings on Purkinje cells uncoupled from climbing fiber input, using a stimulation protocol slightly modified from [12] (Fig 7A). In contrast to the previously performed experiments [10], IPSCs were evoked by stimulation of individual molecular layer interneurons with an ACSF-filled glass pipette positioned in the molecular layer (P2 in Fig 7A). Purkinje cells were held at a membrane potential of -70.5 mV, which was increased to -60.5 mV during stimulation of molecular layer interneurons in order to provide a better signal-to-noise ratio of the evoked IPSCs (eIPSCs; b in Fig  7B). Molecular layer interneurons were stimulated twice (arrows in Fig 7B) with an interstimulus interval of 50 ms to monitor the paired-pulse ratio as a parameter for recording stability and postsynaptic localization of the effect. A pipette solution with low chloride was used in all recordings resulting in an E Cl of -87 mV. Therefore, chloride currents, such as eIPSCs,

PLOS ONE
appeared as outward currents in the recordings (c in Fig 7B). Reduction of eIPSC amplitude was induced, as previously described [12], with five depolarizing pulses to -20.5 mV (inset in Fig 7A) via the patch pipette (P1 in Fig 7A). Each depolarization step lasted for 1 s and was separated from the next depolarization step by another 1 s (0.5 Hz). Glutamatergic inputs as well as CB 1 and GABA B mediated signals were blocked by bath application of CNQX (10 μM), D-AP5 (40 μM), AM 251 (2 μM) and CGP 55845 (2 μM) in all recordings. eIPSCs were completely abolished in the presence of 0.5 μM TTX (Fig 7C) or 20 μM gabazine (Fig 7D). The polarity of the currents together with their sensitivity for TTX and gabazine indicates, that the evoked IPSCs are indeed mediated by molecular layer interneurons.
To investigate whether an ANO2-mediated chloride current attenuates inhibition after depolarization of Purkinje cells, we measured the eIPSC amplitude (red arrow in Fig 7B) every 15 s before and after depolarizing Purkinje cells of wildtype and Ano2 -/mice via the patch pipette ( Fig 8A). To compare recordings of different cells, eIPSC amplitudes were normalized to the averaged eIPSC amplitude preceeding depolarization (baseline). eIPSC amplitudes were significantly reduced in wildtype mice 15 s after depolarization (Fig 8B, upper). In accordance

PLOS ONE
with previous experiments [12], paired-pulse ratios did not change after depolarization (S4 Fig). We observed a complete recovery of eIPSC amplitude with time constant τ = 30.6 s, which was calculated by performing a monoexponential fit to the mean values of eIPSC amplitudes ( Fig 8C). In Ano2 -/mice, the eIPSC amplitude was also reduced significantly after depolarization of Purkinje cells (Fig 8B, lower). Comparing both datasets revealed that the effect of depolarization on the eIPSC amplitudes was significantly stronger in wildtype compared to Ano2 -/mice ( Fig 8D).
The observed reduction of eIPSC amplitudes in wildtype mice occurred without contribution of climbing fibers. Because this reduction was diminished in Ano2 -/mice, we conclude that ANO2 in Purkinje cells must contribute to this process, although other factors appear to be involved as well.
Taken together, our results provide conclusive evidence that ANO2 is expressed in cerebellar Purkinje cells where it reduces simple spike activity during strong activation and at the activation threshold. Additionally, ANO2 seems to be activated upon strong depolarization in the dendrites, which leads to a transient reduction of IPSC amplitudes.

Discussion
One of the main underlying questions of the controversy regarding the role of ANO2 in the olivo-cerebellar system is the exact location of expression of ANO2. As many other ion channels in the brain, ANO2 channels are expressed at low density and, therefore, sensitive methods are needed for their detection [3,6]. Initial evidence for the expression of ANO2 in Purkinje cells has previously been provided with immunohistochemical studies [10,11]. Moreover, with electrophysiological recordings using niflumic acid as a blocker [12] as well as with the Anoctamin-specific blocker T16Ainh-A01 [10], a calcium-dependent chloride current was identified in Purkinje cells. Consistent with these data, we detected the expression of Ano2 mRNA in cerebellar tissue using quantitative real-time PCR (see S1 Fig). These results indicate a more widespread expression Ano2 in the olivo-cerebellar system, not restricted to the inferior olivary nucleus, which has already been established as an expression site of Ano2 [9]. With our in situ hybridization experiments, performed with two different hybridization probes and co-stained with an antibody against the Purkinje cell-specific marker calbindin, we have clearly identified Purkinje cells to express Ano2. Our findings may seem to contradict recent findings obtained from the Ano2-knockout mouse line expressing farnesylated mCherry instead of ANO2 [9]. No mCherry fluorescence was detected in Purkinje cells of this mouse model [9]. Such a discrepancy also occurred when the expression of Ano2 in somatostatin-positive GABAergic neurons of the central lateral amygdala was described [6]. This study also used an in situ hybridization approach to detect an expression of Ano2, although the mCherry-expressing Ano2-knockout mouse line was used for the electrophysiological recordings [6]. This indicates that the mCherry fluorescence might not be strong enough to reliably reveal an expression of Ano2 at low density.
Thus, the current evidence suggests that Ano2 is expressed in two distinct interconnected types of neurons in the olivo-cerebellar system: inferior olivary neurons and Purkinje cells.
In the present study, we monitored changes in excitability of Purkinje cells by analyzing simple spike activity in acute cerebellar slices. To examine the contribution of ANO2-chloride channels, we compared recordings of wildtype to Ano2 -/mice. We did not observe any complex spikes in our recordings, excluding any impact on our results of altered climbing fiber signaling due to the loss of ANO2 in inferior olivary neurons. While spontaneous simple spike discharge was unaltered in Ano2 -/mice, we detected an increased simple spike rate upon threshold activation and during high simple spike activity. These results are supported by the observation that wildtype Purkinje cells develop prolonged interspike intervals compared to Ano2 -/mice, although the initial interspike intervals following current injection are similar. We demonstrated that a chloride current via ANO2 is the main cause of the observed dampening of excitability, as differences in excitability between wildtype and Ano2 -/mice disappeared when the net chloride driving force was minimized. Nevertheless, a small but constant prolongation of the interspike intervals could be observed in wildtype mice following strong activation of Purkinje cells, when the driving force for chloride was reduced. Therefore, a small contribution of a shunt conductance via ANO2 cannot be excluded. The hyperpolarizing impact of ANO2-mediated chloride currents appears to increase the activation threshold of Purkinje cells in wildtype mice. We conclude that ANO2 channels have a certain open probability, even at membrane potentials close to -70 mV, despite their comparatively low sensitivity towards calcium (half-maximally activation at 1.33 μM Ca 2+ [40]) and the high endogenous buffering capacity for calcium of Purkinje cells [41]. We hypothesize that ANO2-chloride channels are located in close proximity to sites of calcium entry, such as voltage-gated calcium channels, as it has been suggested for ANO2 in the hippocampus and the inferior olivary nucleus [3,9]. In the sensitive state of threshold activation even small hyperpolarizing chloride currents via ANO2 seems to be sufficient to significantly lower the activation threshold.
It was previously shown that simple spike activity of Purkinje cells leads to a rise of the intracellular calcium concentration that is largely restricted to the soma [42,43]. Furthermore, the expression of ANO2-chloride channels in the soma of Purkinje cells has previously been suggested with immunohistochemical studies [10,11]. Consequently, increasing somatic calcium concentrations may further activate ANO2 channels in the somatic plasma membrane, thus strengthening the hyperpolarizing chloride current. This chloride current causes only a minimal reduction in simple spike activity during moderate activation, which causes repetitive simple spike activity. Moderate activation in our experiments resembles the state of spontaneous activity (40)(41)(42)(43)(44)(45)(46)(47)(48)(49)(50), dominated by several voltage-gated ion channels. This state is not significantly influenced by the comparatively small chloride current via ANO2 that is induced during moderate activation. With increasing somatic calcium concentrations during stronger activation, the inhibitory effect of ANO2-mediated chloride currents increases and causes a significantly reduced simple spike activity upon strong activation in wildtype mice. In this condition, ANO2-chloride channels are activated more strongly because of the increase of the intracellular calcium concentration and the fact that the channels' sensitivity towards calcium increases with depolarization [44,45]. Thus, our data consistently demonstrate an activitydependent inhibitory influence of ANO2 channels. Additionally, chloride influx via ANO2 increases the afterhyperpolarization upon return to a membrane potential close to -70 mV. Thus, our results are consistent with the hypothesis that ANO2 effects critically depend on the intracellular calcium concentration. While low calcium levels produce an inhibitory effect only near the activation threshold of simple spikes, elevated calcium levels exert a measurable ANO2-mediated effect during strong activation.
Immunohistochemical stainings indicated the expression of ANO2 in the Purkinje cell soma and in the dendritic tree up to the third order branching [10]. A possible expression of ANO2 protein in the fine distal dendrites still needs to be determined. This raises the question as to whether ANO2 is also active when the dendrites are loaded with calcium, which occurs for example after climbing fiber input [43] that triggers dendritic calcium influx [46,47]. Previous studies presented evidence for ANO2-mediated disinhibition following climbing fiber stimulation [10,12]. This effect developed markedly slower than the attenuation of simple spike activity and was proposed to result from chloride accumulation within the dendrite. To identify the role of the intrinsic Purkinje cell ANO2 channels in this disinhibition, we induced calcium influx through somatic depolarization of Purkinje cells via the patch pipette instead via stimulation of climbing fibers. We stimulated molecular layer interneurons in regular intervals to monitor the inhibitory input over time. We observed a reduction of inhibitory postsynaptic currents for several seconds after depolarization of Purkinje cells and revealed a significant contribution of ANO2 to this effect. Therefore, we conclude that ANO2 is also activated in the dendrites and causes local influx of chloride at elevated calcium concentrations. This hyperpolarizing chloride current is accompanied by an accumulation of chloride that has been suggested to reduce the driving force for chloride over the dendritic membrane and, hence, to reduce inhibitory inputs [12]. It still remains to be clarified whether this reduction of inhibition is physiologically relevant or represents a mere side effect of the hyperpolarizing chloride current via ANO2.
In a previous study, hints on an additional expression of ANO1 calcium-activated chloride channels in the soma of Purkinje cells and other GABAergic neurons in the cerebellar cortex were provided using immunohistochemical stainings [10]. The function of these channels in the cerebellum still needs to be determined. A contribution of ANO1 to fast processes such as modulation of spike frequency adaptation of simple spikes appears, however, unlikely due to their activation kinetics in the range of 100 to 300 ms, about ten times slower than the activation of ANO2 [48]. An involvement of ANO1 in the reduction of IPSC amplitude following climbing fiber stimulation, which occurs on much longer time scales, was ruled out using T16Ainh-A01 at a concentration of 5 μM, which causes a reduction of the chloride current via ANO2, but not ANO1 [10]. In the presence of T16Ainh-A01, no reduction of IPSC amplitude could be observed.
The data presented in this study, together with the previously reported evidence for impaired signaling of inferior olivary neurons in Ano2-knockout mice [9], suggest a complex modulation of signal processing in the olivo-cerebellar system via ANO2-chloride channels. In inferior olivary neurons, ANO2 facilitates fast spike generation [9] whereas, in Purkinje cells, ANO2 increases the activation threshold and reduces simple spike activity upon strong activation. This could serve as a mechanism of self-inhibition as it has been reported for ANO2 in fast-spiking thalamocortical neurons [4]. The hyperpolarizing chloride current that follows activation of ANO2 in the dendrites could possibly counteract excitatory inputs from climbing fibers, a hypothesis that has to be tested in future studies when Purkinje cell-specific knockout models for ANO2 become available.
Even though ANO2 seems to affect the excitability of Purkinje cells only mildly, the importance of ANO2 for the olivo-cerebellar system has been shown by the observation that Ano2 -/mice show gait abnormalities and perform poorly in cerebellum-dependent motor-coordination tasks, such as running-wheel and rotarod tests [11]. Moreover, Ano2 -/mice have impaired motor-learning abilities, revealed by the eyeblink conditioning paradigm [9]. Eyeblink conditioning relies on a transient suppression of simple spike activity that seems to be induced by a combination of climbing fiber input and inhibition from molecular layer interneurons [49]. Recently, an increase in intrinsic excitability of Purkinje cells, observed in a mouse line lacking the calcium-gated potassium channel SK2 in Purkinje cells, has been associated with impaired eyeblink conditioning [50]. Moreover, the hyperpolarization after a burst of simple spikes is decreased after eyeblink conditioning [51]. ANO2 is among the factors controlling the eyeblink response, because several of the features contributing to proper eyeblink conditioning appear to be influenced by ANO2-chloride channels. Thus, the behavioral GABAergic neurons in red. ANO2 has been shown to be expressed in inferior olivary neurons, where it accelerates repolarization after high-threshold calcium spikes and, thus, promotes the generation of these calcium spikes [9]. In Purkinje cells, ANO2 attenuates excitability upon strong activation. First indications of an expression of ANO2 in neurons of the deep cerebellar nuclei is presented in Fig 1A, but functional involvement in the cerebellar module awaits investigation. https://doi.org/10.1371/journal.pone.0247801.g009

PLOS ONE
phenotypes of Ano2 -/mice are in line with the notion that ANO2 channels contribute to motion control.
The present study adds evidence to the emerging view that ANO2-mediated chloride currents subserve important functions in modulating neuronal excitability. The results of the present study, combined with the previous results on the inferior olivary nucleus [9], establish for the first time an involvement of ANO2-chloride channels in the modulation of excitability in two different neurons that are directly connected to each other within a neuronal network (Fig 9). Future studies may include the activity of neurons in the deep cerebellar nuclei of Ano2 -/mice, which receive the output signal of Purkinje cells. Such studies may help to better understand this network effect of ANO2 and its implications for cerebellum-dependent motor coordination and learning tasks. As expected, the spontaneous simple spike activity increases in the presence of gabazine (two-way ANOVA: Effect of gabazine: F(1, 144) = 11.34, p < 0.0001; recording condition x gabazine effect interaction: F(3, 144) = 0.73, p = 0.538) (C) The input resistance of the Purkinje cells, which was calculated using a test pulse in the voltage clamp configuration, is comparable between wildtype (black) and Ano2 -/mice (red) and is unaltered by the elevation of E Cl (Kruskal-Wallis test H(3) = 6.126, p = 0.106). (D) The threshold of the first simple spike is comparable between wildtype (black) and Ano2 -/mice (red) for each of the depolarizing current injections and is unaltered by the elevation of E Cl (one-way ANOVA; threshold activation: