Pannexin1 Stabilizes Synaptic Plasticity and Is Needed for Learning

Pannexin 1 (Panx1) represents a class of vertebrate membrane channels, bearing significant sequence homology with the invertebrate gap junction proteins, the innexins and more distant similarities in the membrane topologies and pharmacological sensitivities with gap junction proteins of the connexin family. In the nervous system, cooperation among pannexin channels, adenosine receptors, and KATP channels modulating neuronal excitability via ATP and adenosine has been recognized, but little is known about the significance in vivo. However, the localization of Panx1 at postsynaptic sites in hippocampal neurons and astrocytes in close proximity together with the fundamental role of ATP and adenosine for CNS metabolism and cell signaling underscore the potential relevance of this channel to synaptic plasticity and higher brain functions. Here, we report increased excitability and potently enhanced early and persistent LTP responses in the CA1 region of acute slice preparations from adult Panx1−/− mice. Adenosine application and N-methyl-D-aspartate receptor (NMDAR)-blocking normalized this phenotype, suggesting that absence of Panx1 causes chronic extracellular ATP/adenosine depletion, thus facilitating postsynaptic NMDAR activation. Compensatory transcriptional up-regulation of metabotropic glutamate receptor 4 (grm4) accompanies these adaptive changes. The physiological modification, promoted by loss of Panx1, led to distinct behavioral alterations, enhancing anxiety and impairing object recognition and spatial learning in Panx1−/− mice. We conclude that ATP release through Panx1 channels plays a critical role in maintaining synaptic strength and plasticity in CA1 neurons of the adult hippocampus. This result provides the rationale for in-depth analysis of Panx1 function and adenosine based therapies in CNS disorders.


Introduction
Pannexin1 (Panx1) proteins are integral membrane proteins assembling into large-conductance channels, activated by voltage, ATP, intracellular calcium, stretch, elevated extracellular potassium, or following purinergic receptor activation [1,2,3]. In the central nervous system (CNS) Pannexin1 is expressed in neurons and astrocytes, where it can mediate adenosine 59-triphosphate (ATP) and glutamate release [4,5,6,7]. Independent lines of evidence support a critical role of Panx1 in central nervous system (CNS) pathologies, particularly in epilepsy, stroke, or neuronal cell death [8,9,10,11,12]. In contrast, physiological functions of Panx1 in vivo in the adult CNS are largely uncharacterized. Panx1 is expressed in neurons and astrocytes, where it can mediate adenosine 59-triphosphate (ATP) and glutamate release [4,5,6,7,13]. Since Panx1 is considered to be a major ATP release site, the close anatomical proximity of neurons and astroglia suggests that one of the physiological roles of Panx1 could be in synaptic feedback mechanisms initiated by ATP release. Functional crosstalk between Panx1 and purinergic receptors has been confirmed and ATP regulated ATP release shown [14,15,16,17]. ATP is an agonist of the P2Y and P2X family of purinergic receptors found widely distributed in the CNS in neurons and astrocytes. Purinergic receptor activation by ATP leads to amplification of purinergic signaling thereby affecting synaptic plasticity [18]. Adenosine, a metabolic breakdown product deriving from extra-or intracellular ATP is released from both neuronal and non-neuronal sources. Both ATP and adenosine release depend on a wide variety of stimuli [19] resembling conditions know to open Panx1 channels including response to KCl depolarization, electrical stimuli or glutamate receptor activation [20]. These conditions can create sufficiently high levels of ATP to target purinergic and adenosine receptors at pre-and postsynaptic as well as extrasynaptic sites. In such circumstances modulation of neuronal activities could depend on the spatiotemporal distribution of Panx1, purinergic receptors and the stimulus thus modulating neuronal excitability, synaptic plasticity and coordination of neural networks.
The role of Panx1 and the physiological relevance of this channel need to be determined in vivo. The availability of knock out mice with global inactivation of Panx1 expression provides a unique opportunity to investigate a potential role in synaptic plasticity. In this study, we detected that loss of Panx1 increased excitability and potently enhanced early and persistent LTP responses in the CA1 region of acute slice preparations from adult Panx1 2/2 mice. In line with known effects of extracellular ATP/ adenosine experimental application of adenosine and N-methyl-Daspartate receptor (NMDAR)-blocking normalized this phenotype. This suggests that lack of Panx1 disrupts a feedback mechanism involving pre-and postsynaptic activities. A compensatory transcriptional up-regulation of metabotropic glutamate receptor 4 (grm4) accompanies these adaptive changes, but has to be considered as a secondary effect. Loss of Panx1 promoted distinct behavioral changes, enhancing anxiety and impairing object recognition and spatial learning in Panx1 2/2 mice. We conclude that ATP release through Panx1 channels plays a critical role in maintaining synaptic strength and plasticity in CA1 neurons of the adult hippocampus. This finding adds to the growing body of evidence supporting an important role of Panx1 channels in CNS physiology.

Results and Discussion
Genetic ablation of Panx1 alters postsynaptic responses in hippocampal CA1 area. Conditional knock-out (CMV-Cre/Panx1; Panx1 2/2 ) with full inactivation of the Panx1 gene in the CNS and controls of matching genetic background (Panx1/ LoxP line; Panx1 +/+ ), were used to test a loss-of-function condition [21]. Loss of Panx1 mRNA and protein expression was confirmed (Fig. 1A, B). In adult brains, no structural abnormalities were observed (Figs. S1-S3). Homozygous Panx1 2/2 mice of two founding lines were viable and fertile, but adult animals showed indications for sensitivity to handling stress. Examination of afferent stimulus-dependent field excitatory potential (fEPSP) responses of the Schaffer-collateral CA1 synapse revealed that input-output (IO) relations in adult Panx1 2/2 derived tissues ( Fig. 2A, black circles; n = 18) exhibited distinct responses starting from 25% of the maximum input stimulus power, with evoked responses of the KO significantly shifted towards increased excitability at 25% to 60% of the maximum input stimulus intensity (control, white circles; n = 16, P,0.001; non-parametric analysis, two tailed Mann-Whitney test). Equal sized afferent input stimulus powers (10, 50, 100%) led to distinct and enhanced fEPSP responses in Panx1 2/2 derived slices (Fig. 2B). Broad responses to weak input stimuli and oscillatory activity (Fig. 2B, arrow) at high input powers or following high-frequency stimulation were typical for Panx1 2/2 potential traces.
Pre-and Postsynaptic Components of the Altered Panx1 2/2 LTP Acute pharmacological blocking of Panx1 channels only partially emulated the Panx1 2/2 LTP-phenotype and additional physiological alterations were considered. Since Panx1 is implicated in paracrine and autocrine signaling [35,36], initiated by activity-dependent release of ATP [14,15,37,38,39,40], it was reasonable to speculate that lack of Panx1 in adult mice prompted an ATP misbalance, where upon intracellular postsynaptic ATP levels increase and extracellular ATP, as well as its metabolic breakdown products, fall to critically low levels.
Upregulation of Metabotropic Glutamate Receptor 4 in Panx1 2/2 Mice Next, we investigated whether expression of 84 plasticity-related genes were altered in Panx1 2/2 mice. The result was unexpected since transcriptional alterations were limited to upregulation of metabotropic glutamate receptor 4 (grm4) (Fig. 4A, and Fig. S4). All other candidates showed stable mRNA expression levels (Table  S2). This transcriptional elevation was specific for adult Panx1 2/2 mice, with no alterations found at younger ages (postnatal day 8; data not shown).
To assess cognition, Panx1 2/2 (n = 7) and Panx1 +/+ mice (n = 9) were tested for their ability to discriminate between a known and a new object. In the first trial, animals were exposed to two objects (A) and (B) and allowed to explore them for 5 min. One hour later, the animals were re-exposed to object A and, exposed for the first time, exposed to a new object (C). No significant difference was seen found in the exploration of object A (P = 0.986) or object B (P = 0.968) when the animal groups were compared (Fig. 5B). Similarly, no significant difference was evident in the level of exploration of object A versus object B for either Panx1 +/+ or Panx1 2/2 animals. When 1h-later, exploration of object A was compared with the novel object, C, the Panx1 +/+ animals explored the new object significantly more than the old object ( Fig. 5C; P,0.05). By contrast, Panx1 2/2 mice explored the new object significantly less than the old object (P,0.05), indicating that loss of Panx1 led to deficits in object recognition memory.
Finally, we tested spatial memory capabilities by investigating the ability to remember a place where a treat was hidden during consecutive training trials. Sequential reduction of cookie size and thereby odor cues trained mice to use a memory strategy in preference to olfaction. Both groups performed equally well during initial trials with no significant differences found (n = 9 each group, p.0.05, Student's t-test). At trial 7, both groups found the cookie showing no significant difference, suggesting that olfaction is not impaired in Panx1 2/2 mice (Fig. 5D, Panx1 +/+ = 64621 s, Panx1 2/2 = 45610 s, P = 0.45). Finally, the cookie was removed to challenge the animals' capability to remember the previously trained location. Panx1 +/+ mice remembered significantly better where the treat was hidden (Fig. 5E; path-length to correct location: Panx1 +/+ , 8.461.3 cm; Panx1 2/2 , 12.461.3 cm, P = 0.02). WT mice walking moved less and spent more time searching near the correct (former) cookie location (Movie S1). However, Panx1 2/2 mice were not memory-deficient: they were capable of remembering, to some extent the previous training trials, as their mean path-length to the former location of the cookie (Panx1 2/2 , 12.461.3 cm) was significantly lower compared to untrained littermates (Panx1 +/+ : 18.761.5 cm, P,0.001, Student's t-test). This test suggests that Panx1 2/2 mice are memory-impaired but not memory-deficient.

Discussion
Recent studies have demonstrated that the lack of Panx1 improves the outcome of experimentally induced seizures in juvenile mice [47], but in adults this effect is reversed [48]. This highlights that functions of Panx1 undergo a remarkable change when the brain matures and that distinct age related roles have to be taken under consideration when exploring the physiological and pathological role(s) of Panx1. Here, our study design is based on afferent stimulus-dependent field excitatory potential (fEPSP) recordings of naïve, unchallenged tissue emphasizes the contribution of Panx1 to a defined neuronal network in adult mice. Our results demonstrate that hippocampus-dependent memory is affected in adult Panx1 2/2 mice adding a new twist to the growing number of Panx1 functions. Intriguingly, this effect is accompanied by potently increased LTP. Saturation of LTP is related to impaired learning [49] and object recognition in mice is related to long-term depression [33]. We found that Panx1 2/2 mice have impaired spatial memory and object recognition memory that corresponds well to a counter pro- Figure 5. Behavioral dysfunctions of Panx1 2/2 mice. a, Pre-pulse inhibition of the acoustic startle response (PPI) showing a tendency towards lower levels in Panx1 2/2 (n = 8) compared to Panx1 +/+ (n = 6) mice at the intensities of 62 and 64 dB. At the intensity of 72 dB, a statistically significant reduction of the PPI is detected. b, Object recognition was equivalent in both Panx1 +/+ and Panx1 2/2 mice when the animals were allowed to explored two novel objects (A and B) for 5 minutes. c, One hour later control Panx1 +/+ mice explored the now familiar object A significantly less than the novel object C. By contrast, Panx1 2/2 animals explored object C with significantly less intensity than object A. d, Cookie finding test: Training trials were performed on seven subsequent days, where in trial 1 and 2 a large cookie was used (500 mg), in trial 4 and 5 a smaller cookie (50 mg), and in trial 6 and 7 replaced by a very weak-odorous mouse chow. The time till the mice held the treat in their front paws is depicted in the bar diagram. Panx1 2/2 performed equal to Panx1 +/+ mice in this experiment (P.0.05, Student's t-test) e, A further trial was performed after training trails where no treat was hidden (Movie S1). Bar graphs depict the mean distance of the mice to the location where the cookie was hidden during training trials for the first 60 s. Error bars represent SEM. n = 9 for each mouse group. Panx1 knock out mice have an impaired memory, as they spent less time searching at the correct position (P = 0.02). However, they still remembered the former location to some extent, as they spent significantly more time searching at the correct position as untrained animals (P,0.001, Student's t-test). doi:10.1371/journal.pone.0051767.g005 Panx1 and Synaptic Plasticity PLOS ONE | www.plosone.org ductive propensity towards excessive LTP. Our data support that these Panx1 2/2 mediated effects derive from an increase in neurotransmitter release and subsequent postsynaptic excitability that most likely originates from a long-term depletion in extracellular ATP as shown previously [47], which is accompanied by a compensatory upregulation of grm4 expression in our model. With Panx1 localized in strategic locations in postsynaptic terminals [25] and astrocytes [17,35], our key findings provide compelling evidence for a role of Panx1 in synaptic physiology through signaling induced by ATP release. Our data strongly suggest that Panx1 mediates a feedback response through pre-synaptic activation of A1 receptors and inhibition of glutamate release ensuring that changes in synaptic strength remain within the dynamic range required for effective learning and memory.

Animals
Handling and housing of animals used in this study was performed in compliance with the German Animal Rights law and approved by the Landesamt für Natur, Umwelt und Verbraucherschutz Nordrhein-Westfalen, Germany (Permission No. 50.8735.1 Nr. 100/4). Animals were housed with a 12-hour light/dark cycle and free access to food and water. The generation of Panx1 +/+ mice (Panx1 fl/fl ) with three LoxP consensus sequences integrated into the Panx1 gene, and knockout mice with global loss of Panx1 (Panx1 2/2 ; CMV-Cre/Panx1) have been described [21]. Animals in this study were 3-9 months old.

Reverse Transcriptase-PCR
The Message Sensor TM RT Kit was used to synthesize cDNA from total RNA obtained from dissected tissues. RT-PCR was performed with Panx1 exon4-specific primers. PCR conditions were: 95uC, 15 minutes; 40 cycles at 94uC, 30 seconds; 58uC, 30 seconds; 72uC, 1 minute. Amplified PCR products were analyzed by gel electrophoresis. Forward primer was F 59-CCCTCTGGTCTGCTCTGTGTC-39 and reverse primer was 59-GGGGGTCCAGGTCCGTCTCT-39. The exon4 specific amplicon has a size of 287 bp.

Western Blot
Tissues were dissected, frozen instantly in liquid nitrogen, and homogenized in T-PER buffer lysis buffer (Tissue Protein Extraction Reagent by Thermo Scientific, Inc.) supplemented with complete protease inhibitor (Roche). After debris removal by low speed centrifugation and protein concentration measurements using BCA kit (Pierce), equal amounts of total protein from each sample were resolved on SDS-PAGE gradient 4-12% Bis-Tris gels and transferred to PVDF membrane (both from Invitrogen). To visualize Panx1, homogenates were extracted with 0.1% Triton X-100 and the supernatant supplemented with the loading buffer (Invitrogen). Blots were blocked in 5% milk in Tris-buffered saline (TBS, pH 7.6), probed with the primary antibody overnight, washed in 0.15% Tween20 in TBS, and incubated for 1 h with secondary antibody (1:1,000, Amersham Biosciences, NJ, USA) diluted in TBS. Affinity purified rabbit anti-Panx1 CT-395 (Px-34) antibody was provided by Dr. D.W. Laird (University of Western Ontario, Canada) [22]. A 1:5,000 dilution was used for western blot analysis. Anti-actin antibody was used to control protein loading. Protein bands were visualized using SuperSignal (Pierce) and quantified using the FUJIFILM software.

RNA-expression Profiling
SABiosciences RT 2 First Strand Kits was used for cDNA synthesis from 1 mg total RNA isolated from freshly dissected hippocampi of adult mice (n = 4 for each genotype; 6-9 month old). The kit contains an effective genomic DNA elimination step and a built-in External RNA Control for real-time PCR-based gene expression analysis with SABiosciences' RT 2 Profiler TM PCR Mouse Synaptic Plasticity PCR Array (all from Qiagen Inc., Toronto, Ontario M5J 2T3, Canada). All procedures were performed according to the manufacturers protocol using a DNA Engine Opticon 2 Real-Time Cycler PCR detection system (Bio-Rad Laboratories, CA, USA), with the integrated web-based software package for this PCR Array System was used for all DDC t based fold-change calculations from raw threshold cycle data. Data are presented as Vulcano plot, with thresholds set to: fold difference .1.5, p,0.01 (Student t-test).

Electrophysiology in vitro
Mice were sacrificed by cervical dislocation and horizontal hippocampal slices (350 mm) cut with a Leica VT1000 vibratome (Leica Microsystems, Wetzlar, Germany). Slices were kept in icecold artificial cerebrospinal fluid (ACSF) containing in mM: 124 NaCl, 2.69 KCl, 1,25 KH 2 PO 4 , 2 MgSO 4 , 10 Glucose, 2 CaCl 2 , 26 NaHCO 3 . Slices were incubated for .2 h prior to recording in ACSF at room temperature (RT) for recovery. External solutions were continuously gassed with 95% O 2 /5% CO 2 and applied at a flow rate of 8 ml per minute. During procedures temperature was kept constant at 2360.5uC). Field excitatory postsynaptic potentials (fEPSPs) were recorded by extracellular placement of a metal recording microelectrode (impedances 0, 5-0, 8 MV) (TREC-SE, Multi Channel Systems; Reutlingen, Germany) into the stratum radiatum of the hippocampal CA1 region. A concentric SNEX1200 Wolfram electrode (Hugo Sachs Elektronik, Harvard Apparatus, March-Hugstetten, Germany), was placed into the Schaffer collateral fibers to stimulate hippocampal CA1 pyramidal neurons. The fEPSP responses were driven by bipolar stimuli (50-600 mA) with a STG 1200 (Multichannel systems, Reutlingen, Germany). Before recordings were commenced, input to output correlations were determined to identify the optimum stimulation intensity, which was adjusted to 50% of the evoked maximal response amplitude. To elicit fEPSP response for long-term potentiation (LTP) recordings, 10 min of baseline stimulation at 5 Hz was performed. LTP was evoked by using high frequency stimulation (HFS) with four trains of 10 shocks at 100 Hz every 1 sec. Signals were amplified and filtered by an DAM80 extracellular amplifier (World Precision Instruments, Sarasota FL, USA), digitized at 20 kHZ and displayed, stored and analyzed using WinWCP software (Strathclyde; Biologic, Knoxville TN, USA). A 16bit analog to digital converter (BNC 2110 connected to Ni-PCi 6229; National Instruments; Munich, Germany) was used to digitize the signals.
Electophysiological Data Analysis fEPSP data were normalized as a percentage of control, based on average amplitudes from the 10 min recording immediately before LTP protocol application. For each time point, consecutive responses at 20 sec intervals were averaged and the results were expressed as the mean percentage 6 standard error of the mean (s.e.m.) and summarized in table S1. LTP responses were analyzed as described [26]. All values are expressed as mean 6 standard error of the mean (s.e.m.). The level of significance was set at p,0.01 [p,0.01 = *; p,0.001 = **; p,0.0001 = ***]. fEPSP amplitude data, summarized in Box-and whisker-plots in figures 1,2,3 are represented as sample minimum, lower quartile, median, sample maximum, upper quartile and considered outliers.

Prepulse Inhibition (PPI) of the Acoustic Startle Response
All animals were subjected to 4 behavioral test sessions investigating PPI of ASR as described previously [27]. Prior to experiments animals had been individually handled and habituated to the test apparatus [28]. PPI of the ASR was measured in a sound-attenuated isolation chamber (41641641 cm) using a movement-sensitive piezoelectric measuring platform connected to a personal computer with an analogue to digital (AD) converter (Startle Response System, TSE, Bad Homburg, Germany) [27,29]. During test sessions, animals were placed in a wire mesh cage (22.5 cm68 cm68.5 cm) mounted on the transducer-platform. For acoustic stimulation, two loudspeakers were used, mounted on both sides of the test cage at a distance of 4 cm. On the day of PPI testing, the animals were transported to the startle-box room and left undisturbed for 30 min. The experiment consisted of a 5 min acclimatization phase and a test session. During the acclimatization time, animals received background noise (60 dB sound pressure level (SPL), white noise) followed by 10 initial startle stimuli (100 dB SPL, white noise) lasting each for 20 ms (0 ms rise/fall times). The test session consisted of seven different trial types given in a pseudorandom order: (1) pulse alone (100 dB SPL white noise, 20 ms duration); (2) control (no stimulus); (3)(4) prepulse alone (72 or 68 dB, pure tone, 10 Hz, 20 ms duration); (5-7) pre-pulse (72, 68, or 64 dB) each followed by a pulse with an interstimulus interval of 100 ms. A total of 10 presentations of each type was given with an inter-trial interval randomized between 20,000 ms and 30,000 ms [30]. Background noise intensity during the whole experiment was 60 dB SPL. The entire test session took about 40 min.
PPI was calculated according to the formula 100-100% 3 (PPx/ PA), in which PPx is the mean acoustic startle response (ASR) of the 10 PPI trials (separate for each individual pre-pulse intensity) and PA is the mean ASR to the pulse alone trials [28]. Analysis of variance (ANOVA) was used for all comparisons (treatment, prepulse-intensity). A probability level (P) of less than 0.05 was considered statistically significant.

Object Recognition Test
In a chamber (40640640 cm) that is familiar to the animals, two novel objects (i.e. A and B) were presented for 5 min. After a delay of 60 min, one familiar and one novel object (i.e. A and C) were presented to test for object recognition memory. The presentation of objects lasted for 5 min, where the animals were left to explore freely, and were removed from the recording chamber after the presentation.
The objects and the recording chambers were cleaned thoroughly between task trials to ensure the absence of olfactory cues. The objects were distinctly different from one another and heavy so that the mice could not move them. Several copies of each object were available.
Behavioral data were recorded from cameras positioned above the chambers, and digitally stored. Exploration of the objects was then analyzed post-hoc using the within-object area scoring system, which was defined as sniffing of the object (with nose contact or head directed to the object) within ,2 cm radius of the object [31]. Standing, sitting or leaning on the object was not scored as object exploration. OR data were expressed as a percentage of the total exploration time for each object per experiment. [32,33]. The results across animals were expressed in terms of mean 6 s.e.m. The data were then statistically assessed using the Student's t-test by comparing group means with the fixed value of 50%, which represents no differentiation between objects. The significance level was set at p,0.05 [32].

Cookie Finding Test
Mice were trained in cages marked with letter ''X'' as visual cue. In trials 1-3 a cookie was buried beneath ,6 cm (400 g) of woodchip bedding in their home cage. Reducing the cookie size in trials 4 and 5 reduced the olfactory component in this test. Finally, in trials 6 and 7 weakly odorous mouse chow replaced the original cookie. Training was performed on subsequent days and the treat was always hidden at the same position, with animals allowed to locate the treat in a 10 min. period. The time till the mice held the treat in their front paws was defined as finding time. After seven training trials animals were exposed to the test trial with no hidden treat. The movements were recorded and tracked with the EthoVision XT7 Software from Noldus (Wageningen, Netherland). The walking distances to the place where the cookie was hidden during the training trials were used to quantify the spatial memory of the mice. Statistical significance was tested by Student's T-test. The significance level was set at P,0.05. Figure S1 Morphology of adult Panx1 +/+ and Panx1 2/2 mouse brains. Photographs of fixed 9 month-old male control (Panx1 +/+ , left panels) and Panx1 2/2 (right panels) brains showing no macroscopic differences in dorsal (A, B) and ventral views (C, D). Scale Bars A-D = 5 mm. (JPG) Figure S2 Comparison of frontal brain sections Panx1 +/+ and Panx1 2/2 mouse brains. (A, B) Overview representing slices (1.5 mm) at the level of the dorsal hippocampus reveals no obvious regional differences between cortex (Cx), hippocampus, thalamus, (Th), hypothalamus (Hy) and amygdala (Amy) between Panx1 +/+ (left panels) and Panx1 2/2 mice (right panels). All animals were 9 months old. Table S1 Summary of the mean 6 SEM values for Panx1 +/+ and Panx1 2/2 derived early phase LTP and late phase LTP. Data are listed according to the bath applied pharmacological treatment. Note, that wash in of pharmacology was performed at least 10 min in advance to measurements and was kept upright during the whole phase of LTP recordings. P-Values reveal significances for Holm-Sidack post hoc comparisons, which were performed following one-way ANOVA analyses. ns = non-significant (DOCX) Methods S1 This section describes methods used to obtain the data described in the supporting information section. (DOC)

Supporting Information
Movie S1 Typical behavior of trained control (Panx1 +/+ ; cage on the left) and knock-out (Panx1 2/2 ; cage on the right) mice in cookie finding assay. The ''X'' presented in the lower left corner signifies where the cookie was stored during training sessions. Control mice immediately try to find the cookie in this corner, and when unsuccessful choose the top right hand corner instead. The Panx1 2/2 explores the cage, however, lack indications of a systematic, memory-based exploration strategy. (MOV)