KCC2 regulates neuronal excitability and hippocampal activity via interaction with Task-3 channels

The K+/Cl− co-transporter KCC2 (SLC12A5) regulates neuronal transmembrane chloride gradients and thereby controls GABA signaling in the brain. KCC2 downregulation is observed in several neurological and psychiatric disorders including epilepsy, neuropathic pain and autism spectrum disorders. Paradoxical, excitatory GABA signaling is usually assumed to contribute to abnormal network activity underlying the pathology. We tested this hypothesis and explored the functional impact of chronic KCC2 downregulation in the rat dentate gyrus. Although the reversal potential of GABAA receptor currents was depolarized in KCC2 knockdown neurons, this shift was fully compensated by depolarization of their resting membrane potential. This effect was due to downregulation of Task-3 leak potassium channels that we show require KCC2 for membrane trafficking. Increased neuronal excitability upon KCC2 suppression altered dentate gyrus rhythmogenesis that could be normalized by chemogenetic hyperpolarization. Our data reveal KCC2 downregulation engages complex synaptic and cellular alterations beyond GABA signaling that concur to perturb network activity, thus offering novel targets for therapeutic intervention.


Introduction
Fast synaptic inhibition in the brain is primarily mediated by GABAA receptors (GABAAR), which are ligand-gated receptors associated with an anion-permeable conductance. GABAAR currents are carried by fluxes of chloride and, to a lesser extent, bicarbonate ions 1 .
Consequently, the polarity of the net ion flux through GABAAR channels relies on both the transmembrane gradients of these ions as well as the neuronal resting membrane potential, which thereby determine the driving force of GABAAR currents. In mature neurons, transmembrane chloride gradients are predominantly regulated by the opposing actions of the two cation-chloride co-transporters (CCCs) KCC2 and NKCC1, which respectively mediate outward and inward co-transport of chloride and potassium ions 2 . Thus, postnatal upregulation of KCC2 expression is associated with a progressive hyperpolarizing shift in the reversal potential of GABAAR-mediated currents (E GABA ) 3 .
In mature neurons, KCC2 expression and function is rapidly regulated by neuronal activity via multiple posttranslational mechanisms 4,5 . Thus, Ca 2+ influx through postsynaptic NMDA receptors or during prolonged postsynaptic firing 6 rapidly reduces KCC2 membrane expression and function through protein phosphatase 1-dependent dephosphorylation of its Ser940 residue and protein cleavage by the calcium-activated protease calpain 7,8,9 . Conversely, chloride influx through GABAA receptors stabilizes KCC2 at the plasma membrane via chloride-mediated inhibition of the serine/threonine WNK1 kinase and its downstream effectors SPAK/OSR1, which phosphorylate KCC2 on Thr906 and Thr1007 residues 10,11,12 . Finally, KCC2 expression is also regulated by several neuromodulators acting on G-protein couples receptors 13 as well as neurotrophins such as BDNF acting via TrkB signaling 14 .
KCC2 dysregulation has been associated with numerous neurological and psychiatric disorders involving synaptic alterations 15,16 . These include epilepsy 17,18,19,20,21,22,23 , neuropathic pain 24,25 , posttraumatic spasticity 26 , Huntington disease 27 , schizophrenia 28 and Rett syndrome 29,30,31 . In most cases, reduced KCC2 expression was associated with a depolarizing shift in E GABA that could be partly reversed using NKCC1 antagonists or drugs acting to enhance KCC2 function 18,25,27,30,32 . Since these drugs also ameliorated the pathological symptoms, those were generally assumed to primarily reflect a defect in neuronal chloride transport and altered GABAergic neurotransmission.
However, KCC2 functions in neurons extend beyond the mere control of chloride transport.
Through interactions with multiple transmembrane and intracellular partners, KCC2 was shown to regulate dendritic spine morphology 33,34 and actin cytoskeleton 35,36 , as well as the strength and long-term plasticity of glutamatergic synapses 33,35 . Recent functional proteomics data revealed additional putative KCC2 partners, including some involved in the recycling and trafficking of various transmembrane proteins and receptors 37 . Those may then either influence KCC2 function or, conversely, be regulated by KCC2. Thus, dysregulation of KCC2 expression may affect a variety of neuronal intrinsic and synaptic properties and thereby contribute to abnormal activity patterns that underlie neurological and psychiatric disorders.
Deciphering the relative contribution of the various perturbations upon KCC2 downregulation may then help predict the most effective rescue strategies.
In the present study, we therefore explored the functional impact of chronic KCC2 downregulation in the rat dentate gyrus at the cellular, synaptic and network activity levels.
We found that KCC2 knockdown had no significant effect on the driving force of GABAAR currents at rest but instead enhanced neuronal excitability and EPSP/spike coupling through membrane depolarization and increased resistance. Pharmacological and biochemical evidence suggest this reflected downregulation of a leak potassium conductance carried by TWIK-related acid-sensitive 3 (Task-3) channels, the expression of which was regulated via interaction with KCC2. KCC2 knockdown was not sufficient to promote epileptiform activity but altered dentate gyrus network activity, which was normalized by restoring granule cell membrane properties. Our results show that KCC2 downregulation affects network activity primarily through enhanced neuronal excitability, suggesting novel targets for therapeutic intervention.
The shift in E GABA and V rest was of similar magnitude such that the driving force of ion flux through GABAARs at rest was unaffected by KCC2 knockdown ( Fig S1). Therefore, KCC2 suppression in dentate granule cells does not affect steady-state GABAergic transmission, due to an unexpected depolarizing shift in resting membrane potential.
Depolarized membrane potential and increased input resistance likely reflect a reduction in potassium currents operating at rest. We tested this hypothesis by recording potassium currents upon voltage ramps from -120 to -50 mV at 0.03 mV/ms while blocking synaptic transmission. In KCC2 knockdown granule cells, potassium conductance was reduced throughout the range of potentials tested (Fig 2G-H). This reduction however was of higher magnitude for outward than inward currents, as evidenced by reduced rectification index (I -60 / I -120 ) in KCC2 knockdown vs control granule cells (0.46 ±0.05 vs 0.82 ± 0.08, t-test, t 37 = 3.62, p = 0.0009, Fig 2I). This suggests that outwardly rectifying leak potassium channels might be down-regulated upon KCC2 suppression. Two-pore-domain potassium channels (K2P) are a large family of leak potassium channels comprising both inward and outward rectifiers 39,40 . We therefore tested whether currents carried by K2P channels may be reduced in KCC2 knockdown neurons. Bath application of bupivacaine (100µM), a non-selective K2P channel antagonist significantly reduced potassium currents in neurons expressing nontarget shRNA but had very little effect in neurons expressing KCC2-directed shRNA ( Fig 2G).
Trek-2 and Task-3 are the most prominent outward-rectifier K2P in dentate gyrus granule cells 41,42,43 . Immunohistofluorescence imaging of Trek-2, however, revealed a predominant expression in mossy fibers ( Fig S4A) with little or no expression in somato-dendritic compartments, suggesting Trek-2 may not prominently contribute to granule cell resting membrane potential and membrane conductance. Conversely, Task-3 expression was primarily observed in granule cell somata and dendrites (Fig S4B). We therefore examined whether Task-3 protein expression in granule cells might be affected by chronic KCC2 suppression. Immunofluorescence confocal imaging indeed revealed a reduced expression of Task-3 in granule cells expressing KCC2-directed shRNA compared to non-target shRNA ( Fig 3A). We then investigated the mechanisms underlying Task-3 down-regulation upon KCC2 silencing. KCC2 has been shown to interact with a variety of proteins, including transmembrane proteins such as GluK2 44 , Neto2 45 and GABA B receptors 46 that regulate KCC2 membrane expression or the intracellular guanylyl exchange factor β PIX which is regulated through direct interaction with KCC2 35,36 . We therefore asked whether KCC2 might also interact with Task-3 channels and thereby influence their membrane expression and/or function. In order to test whether KCC2 interacts with Task-3 channels in neurons, we performed a co-immunoprecipitation assay from adult rat hippocampal homogenates. We observed that anti-Task3 antibodies pulled down KCC2, indicating an interaction between the endogenous proteins in hippocampal neurons ( Fig 3B). In order to characterize KCC2-Task-3 interaction further, we co-immunoprecipitated exogenously expressed Flag-tagged KCC2 and HA-tagged Task-3 in HEK293T cells. In lysates from double-transfected cells, anti-HA antibodies pulled down Flag-KCC2 together with HA-Task-3 while anti-Flag antibodies pulled down HA-Task-3 and Flag-KCC2 ( Fig 3C). Importantly, this interaction was specific for Task-3 since overexpressed HA-tagged GFP failed to interact with KCC2 in a similar assay (Fig.   3C). We then asked whether interaction with KCC2 may influence the expression of Task-3 channels. To test this hypothesis, we expressed HA-Task-3 in Neuro2A cells either alone or together with Flag-KCC2 and analyzed Task-3 expression by immunostaining and confocal imaging. In the absence of KCC2, Task-3 expression at plasma membrane was reduced and accumulated in an intracellular compartment ( Fig. 3D-E). Line scans of immunofluorescence across the cell membrane and cytoplasm therefore revealed a reduced membrane to intracellular intensity ratio in cells lacking KCC2 (0.74 ± 0.08 vs. 2.47 ± 0.34, Mann-Whitney, p = 1.63.10 -7 , Fig 3E). Together, our results show that KCC2 interacts with Task-3 and regulates its membrane expression in dentate gyrus granule cells, thereby regulating their intrinsic excitability. We next asked how reduced leak conductance and enhanced excitability upon KCC2 suppression might affect EPSP/spike coupling in granule cells. Previous work showed that loss of KCC2 expression in hippocampal neurons led to reduced synaptic strength at glutamatergic inputs due to enhanced lateral diffusion of AMPA receptors 33 . Enhanced intrinsic excitability may then act as a homeostatic compensatory mechanism for reduced synaptic excitation 47 . To test this hypothesis, we first compared synaptic strength at glutamatergic inputs onto granule cells by recording miniature excitatory postsynaptic currents (mEPSCs). As previously reported, KCC2 knockdown granule cells showed a ~20% decrease in mEPSC amplitude (5.06 ± 0.43 vs 6.54 ± 0.28 pA, Mann-Whitney, p = 0.0009, Perforant path input to the dentate gyrus provides synaptic excitation to granule cells as well as feedforward inhibition that contributes to their sparse activation by entorhinal afferents 48 . In order to explore EPSP/spike coupling of perforant path inputs in a more physiological setting, we recorded evoked fEPSP and population spike upon stimulation of perforant path inputs. The recording electrode was positioned in a densely infected area of the granular layer and stimulus intensity was set such that the population spike amplitude was about half its maximum amplitude (see Methods). In slices infected with lentiviruses expressing KCC2directed shRNA, fEPSP/population spike coupling was increased by ~47% compared to control slices expressing non-target shRNA (2.00 ± 0.25 vs 1.36 ± 0.15 ms -1 , t-test, t 29 = 2.18, p = 0.0374, Fig 4F). This result further supports that increased excitability upon KCC2 knockdown is not homeostatic but instead acts to promote granule cell recruitment by entorhinal afferents. Importantly, bath application of the GABAAR antagonist picrotoxin (100µM) increased the amplitude of population spike similarly in slices expressing either nontarget shRNA (1.12 ± 0.15 mV vs 0.88 ± 0.13 mV, paired t-test, t 13 = 3.33, p = 0.0054) or KCC2-directed shRNA (1.21 ± 0.17 mV vs 0.79 ± 0.11 mV, paired t-test, t 15 = 4.29, p = 0.0006; Fig 4G). These results suggest that GABAergic transmission, although depolarizing in dentate granule cells 49 (Fig. 1E), remains shunting and inhibitory, independent of KCC2 expression. Therefore, KCC2 suppression in dentate granule cells promotes their recruitment by entorhinal afferents primarily through enhanced excitability and EPSP/spike coupling rather than altered GABA signaling.
The dentate gyrus is often considered as a filter or a gate for activity propagation from the entorhinal cortex to the hippocampus 50 , in particular owing to dentate granule cells' sparse firing. These properties allow for pattern separation 51 and may prevent runaway excitation in the hippocampus 52 . Enhanced EPSP/spike coupling in dentate granule cells upon KCC2 suppression may then act to increase their excitatory drive by entorhinal afferents and thereby alter hippocampal rhythmogenesis. We explored this hypothesis using chronic intrahippocampal recordings from rats injected with lentivirus expressing either KCC2directed or non-target shRNA (n=7 and 5, respectively, Fig 5A-B). In these animals, about 40-70% granule cells were infected over 1-2 mm around the injection site ( Fig. 5B and Fig   S5). Unlike a recent study using a similar approach 53 , we did not observe spontaneous seizures in any of these animals. Susceptibility to pilocarpine-induced seizures was also unaffected by KCC2 knockdown in the dorsal dentate gyrus (Fig S6A-C). Chronic, intrahippocampal recordings using linear silicon probes did not reveal interictal spikes or fast ripples (Fig S6D-E) that could represent hallmarks of an epileptic hippocampal network even in the absence of behavioral symptoms 54 . These results indicate that focal KCC2 suppression in the dentate gyrus is not sufficient to trigger or promote epileptiform activity.
Further analysis of hippocampal rhythmogenesis during various behavioral states however revealed significant alterations upon KCC2 knockdown in the dentate gyrus. Thus, while theta-band activity recorded during exploration and rapid-eye movement (REM) sleep was not significantly affected ( Fig S7), we observed a specific increase in both the amplitude (Kolmogorov-Smirnov, p < 0.001, Fig 5F) and frequency (Kolmogorov-Smirnov, p < 0.001, Fig 5E) of dentate spikes during slow-wave sleep (SWS). Dentate spikes are sharp hilar LFP events associated with rest 55 that arise from a transient increase in the excitatory drive from the entorhinal cortex, as revealed by current source density analysis 55 (Fig. 5D-E). Our results suggest these changes in dentate spike activity resulted primarily from increased excitability of dentate granule cells. Restoring granule cell excitability independently of chloride export and GABA signaling was then predicted to rescue dentate spike activity. We tested this prediction by co-injecting animals with adeno-associated viruses (AAV2.1) expressing the inhibitory receptor hM4D(Gi) 56 and either non-target or KCC2-directed shRNA with GFP. Using this approach, all granule cells expressing GFP also expressed hM4D(Gi), thereby enabling selective silencing of neurons expressing shRNA sequences ( Fig. 5G-H).
We first verified that hM4D(Gi) activation was sufficient to hyperpolarize transduced granule cells in vitro. Indeed, in whole cell recordings from transduced neurons, bath application of the hM4D(Gi) ligand CNO (10µM) induced a hyperpolarizing shift of V rest by -5.66 ± 1.42 mV (paired t-test, t 5 = 4.36, p = 0.0073, Fig S8). As in lentivirus-infected animals, AAV-mediated KCC2 suppression in the dorsal dentate gyrus increased the amplitude and reduces the frequency of dentate spikes during SWS (n = 4 rats infected with non-target shRNA and 3 with KCC2-directed shRNA, Kolmogorov-Smirnov, p < 0.001, Fig 5I). Following systemic injection of CNO (1 mg/kg, i.p.), the amplitude of dentate spikes in animals injected with KCC2-directed shRNA was reduced to levels below those of saline-treated control animals ( Fig 5I). Similarly, CNO injection reduced the frequency of dentate spikes in both KCC2knockdown and control animals (Kolmogorov-Smirnov, p < 0.001, Fig 5I). Together, our data demonstrate that KCC2 suppression in dentate granule cells leads to altered dentate gyrus activity through a reduced Task3-mediated potassium conductance and a subsequent increase in neuronal excitability, independent of GABA signaling.

Discussion
Cation-chloride cotransporters (CCCs) are essential to neuronal chloride homeostasis. Thus, changes in their expression level in the developing brain are associated with changes in the polarity of GABA transmission 3 and altered CCC function or expression is observed in a variety of neurological and psychiatric conditions 15,16 . In the pathology, KCC2 downregulation is often assumed to primarily affect GABA signaling and thereby promote hyperexcitability 17,18,25,26,27,57 . However, this hypothesis was not tested in isolation from other, potentially confounding pathological factors. Here we directly tested the impact of a chronic KCC2 suppression both on the cellular and synaptic properties of hippocampal neurons as well as on network activity associated with behavior. We demonstrate that KCC2 knockdown in dentate granule cells has little effect on GABA signaling at rest but instead increases their excitability via interaction with the leak potassium channel Task-3. We establish that this increased excitability strengthens EPSP/spike coupling in dentate gyrus granule cells thereby acting to enhance their recruitment by entorhinal afferents. This in turn results in altered local  Task-3 membrane targeting requires KCC2 expression in heterologous cells. Thus, in cells lacking KCC2, Task-3 was sequestered intracellularly and its membrane expression was reduced ( Fig. 3). We did not attempt to fully identify the molecular determinants of this interaction or the mechanism involved in KCC2-dependent trafficking of Task-3 channels. However, it is remarkable that whereas Task-3 was not found among KCC2 interactors by functional proteomics, several members of the YWHA family encoding 14-3-3 proteins were identified as putative partners 37 . 14-3-3 are adaptor proteins involved in a variety of biological functions that include masking phosphorylated or interaction sites of their targets or altering their subcellular localization 58 . In particular, 14-3-3 proteins were shown to bind to and influence the membrane traffic of both Task-1 and Task-3 channels 59 . This effect is mediated through a KRR retention signal in the carboxy-terminal domain of the channels that is masked by 14-3-3 binding. Thus, phosphorylation-dependent binding of 14-3-3 suppresses coatomer protein I (COPI)-mediated retention of Task channels in the endoplasmic reticulum 59,60 . We therefore propose that KCC2 binding to both 14-3-3 and Task-3 may somehow influence the interaction between the two proteins, thereby preventing COPI-mediated intracellular sequestration of the channels. We cannot exclude, however, that KCC2-Task-3 interaction may also occur at the plasma membrane and act to stabilize Task-3 channels.
In mature neurons, KCC2 plays a prominent role in maintaining transmembrane chloride gradients, thereby influencing ion flux through the chloride-permeable GABAA receptor channels. Thus, KCC2 pharmacological blockade 45,46 , genetic ablation 61,62 or suppression 63 , or activity-dependent modulation 6,11,46,64 all result in a depolarization of the reversal potential of GABAAR-mediated currents (E GABA ). Consistent with these observations, we found that chronic KCC2 knockdown in dentate granule cells resulted in a depolarizing shift of E GABA by about 9 mV. This effect was, however, almost fully compensated by a depolarization of their resting membrane potential (V rest ) of about the same magnitude ( Fig. 1), resulting in a driving force of GABAAR currents that was virtually unaffected at rest. This effect was not specific of dentate gyrus granule cells as it was observed also in CA1 pyramidal neurons (Fig. S3).
Despite the importance of the membrane potential in setting the efficacy and polarity of GABA signaling, only one study investigated the effect of KCC2 suppression on V rest in cerebellar neurons 61 . Thus, conditional KCC2 ablation resulted in a depolarization of V rest that also compensated the shift in E GABA in granule cells but not Purkinje cells. Interestingly, hippocampal and cerebellar granule cells as well as CA1 pyramidal cells strongly express Task-3 channels 42,65 whereas in Purkinje cells predominantly express Task-1 and Task-5 Thus, differential expression of leak potassium channels may underlie cell-specific effects of KCC2 knockdown on resting membrane potential and the driving force of GABAAR currents.
KCC2 downregulation has often been linked to epileptogenesis. Thus, KCC2 expression is reduced in several animal models of experimental epilepsy including kindling 67 (Fig. 4C-D). Several arguments support that these effects were mostly independent of KCC2-induced alterations in GABA transmission. First, GABA transmission was still inhibitory upon KCC2 suppression, as indicated by the unchanged driving force of GABAAR-mediated currents (Fig. 1) and the excitatory effect of picrotoxin on evoked population spikes (Fig. 4G). Second, increased dentate spike amplitude was fully compensated by chemogenetic hyperpolarization of KCC2 knockdown neurons (Fig. 5). Since GABAAR activation is depolarizing in dentate granule cells 49 , hM4D(Gi)-induced hyperpolarization rescued V rest in infected granule cells but further enhanced the depolarizing nature of GABA transmission. If GABA signaling was involved in the increased dentate spike amplitude and frequency, hM4D(Gi) silencing would then be expected to worsen, not rescue this phenotype.
Together, our work reveals for the first time the effects of a chronic downregulation of KCC2 expression in an intact network. We show these effects are diverse but primarily involve increased neuronal excitability, not reversed GABA signaling. They lead to altered network activity without however promoting seizures. Thus, KCC2 suppression in the dentate gyrus primarily affects dentate spike activity with no detectable effect on theta-band activity throughout the hippocampal formation (Fig. S6). This rather subtle phenotype may still be of physiopathological relevance for disorders associated with KCC2 downregulation. Dentate spikes coincide with periods of stronger gamma-band activity coherence between hippocampus and cortex and are temporally correlated to the power of sharp-wave ripples 77 .
Since both dentate spikes 78 and sharp-wave ripples 79 have been associated with memory consolidation in hippocampal-dependent learning, we predict KCC2 suppression in the pathology may impact cognitive performances through a combination of GABA-dependent and -independent mechanisms. Our results suggest therapeutic strategies aiming to either restore neuronal chloride homeostasis by blocking NKCC1 function 57 and neuronal excitability by targeting leak potassium channels 80 , or to stabilize KCC2 membrane expression 25,32 may best compensate for altered network activity and associated behavioral or cognitive deficits.

Animals
Juvenile male Wistar rats were obtained from Janvier Labs (Le Genest Saint Isle, France).
All procedures conformed to the International Guidelines on the ethical use of animals, the French Agriculture and Forestry Ministry guidelines for handling animals (decree 87849, licence A 75-05-22) and were approved by the Charles Darwin ethical committee (APAFIS2015111011588776v3).

Virus and constructs
Rat slc12a5-specific and non-target shRNA sequences 33 were introduced in pTrip vector under U6 promoter and used to produce purified lentiviral particles (titer 7-9 x 10 9 TU/ml, UNC Vector Core Facility) 35 . For some experiments, these same sequences were introduced in pAAV vector under U6 promoter and used to produce purified AAV particles (AAV2.1-shRNA-GFP, titer 10 13 TU/ml, Atlantic Gene Therapy, Nantes). AAV particles expressing inhibitory DREADD under synapsin promoter were obtained from the Viral Vector Facility of Zurich University (AAV2.1-hM4D(Gi)-mCherry, titer 10 13 TU/ml). For pilocarpine experiments, animals were also implanted with EcoG electrodes soldered to stainless steel screws. Two electrodes were placed above each hippocampus, at the same coordinates as the injection sites. A third screw was placed above the cerebellum and used as reference.

Intracerebral probe implantation
For intra-hippocampal recordings, animals were anesthetized with isoflurane one week following viral injection, and injected subcutaneously with 0.05 mg/kg of buprenorphine for analgesia. One 16-channels silicon probe was implanted in the right hemisphere, at the site of the injection. Signals were monitored during the insertion of the probe to ensure correct localization of the probe in the hippocampus. One stainless steel screw was placed above the cerebellum and connected to the probe as a reference. Four additional screws were placed above the prefrontal cortex and the contralateral hemisphere to secure the dental cement adhesion. During the recovery period (minimum 3 days), rats were daily monitored and additional injection of buprenorphine (subcutaneous, 0.05 mg/kg) were made in case of detectable pain.

Slice preparation
10 to 20 days after viral injection, animals were deeply anesthetized with ketamine / xylazine  Resting membrane potential was measured using I=0 current clamp mode one minute after break-in. Input resistance was measured in current-clamp mode with -50 pA current steps.
Input/output curves were derived from series of 800 ms current steps (ranging from -100 to

Extracellular field recordings
For extracellular recordings, a recording pipette of 2-3 MΩ resistance was filled with ACSF and inserted within the granular layer, in a densely infected area of the dentate gyrus. A tungsten bipolar electrode (0.5 MΩ) was used to stimulate perforant path inputs. Stimulation intensity was adjusted to induce a population spike of about half its maximum amplitude.
Field EPSP (fEPSP) slope was determined over a 1ms window preceding the population spike. Population spike amplitude was assessed as the distance between its peak and the baseline extrapolated from its initiation and termination points using a routine written under Matlab (The MathWorks, Inc., Natick, MA, USA).

Co-immunoprecipitation assays from rat hippocampal homogenates
Hippocampi were dissected from adult Sprague Dawley female rats and homogenized by sonication in co-immunoprecipitation buffer as above (1 ml/100 mg of tissue) and solubilized by rotation for 2 to 4 hours at 4°C. After centrifugation at 20,000 g for 40 minutes at 4°C, protein content of the supernatant was measured by BCA protein assay (Pierce). 6.5 mg of proteins were transferred into microtube and 5 μ g of primary antibody (Rabbit anti-KCC2, Millipore, and Rabbit anti-Task3, Alomone) or nonimmune IgG control antibody of the same species (Jackson Immunoresearch) were added. Samples were then incubated overnight at 4°C under rotation and processed as above.

Co-immunoprecipitation assays from HEK293T cells
HEK293T cells were grown in DMEM GlutaMAX (Invitrogen) supplemented with 5 g/l glucose and 10% fetal bovine serum. Cells (60-70% confluent) were co-transfected using transfectin Complexes were precipitated with 20 µl protein-G magnetic beads (Invitrogen) for 2 hours at 4°C. Beads were then washed twice with 1ml of co-immunoprecipitation buffer and once with triton-free co-immunoprecipitation buffer. Bound complexes were eluted in 4X sample buffer (Invitrogen) for 1 hour at 37°C. Samples were analyzed by SDS-PAGE and western blot.

Western blotting
Proteins were separated on a 4-12% SDS polyacrylamide gradient gel (Invitrogen) and transferred onto a nitrocellulose membrane (GE Healthcare). For co-immunoprecipitation assays from rat hippocampal homogenate, blots were probed with antibodies against KCC2 (1/1000, Millipore) and Task3 (1/250, Santa Cruz). For co-immunoprecipitation assays from HEK293T cells, blots were probed with HA (1/3000, Cell signaling) and KCC2 or FlagM2 For images of the entire dentate gyrus, images were acquired using a 10X objective and stacks of 10 to 20 µm optical sections were acquired at a pixel resolution of 1024 x 1024 with a z-step of 1 µm.

Behavior and recordings
All recordings took place in a dimly-lit area enclosed by black curtains. Animals were handled for at least 3 days for habituation and to ensure stability of the recordings ( minutes before starting recording. All recordings took place less than 3 hours after i.p. injection. Data were acquired at 20 kHz using an Intan recording controller (Intan Technologies, Los Angeles, USA) and Intan Recording Controller software (version 2.05).

Data analysis
All analyses were performed offline using Matlab built-in functions and custom-written scripts as well as Chronux (http://chronux.org/) and the FMAToolbox (http://fmatoolbox.sourceforge.net/). Power spectra and spectrograms were computed using multi-tapers estimates on the raw LFP. Theta power was determined in the 5-10 Hz band.
Dentate spikes detection was performed after high-pass filtering (> 30Hz), squaring and normalizing the field potential in the dentate gyrus. DS were defined as events of positive deflection peaking at > 8 s.d. and remaining above 4 s.d. for less than 30 ms. Current source density analysis (CSD) revealed two distinct DS profiles, previously described as DS1 and DS2 55 . Since most (70-80 %) DS in our recording conditions were of type 2, we did not further attempt to distinguish between DS subtypes in our analysis.
Ripple detection was performed by band-pass filtering (100-600 Hz), squaring and normalizing, followed by thresholding of the field potential recorded in CA1 pyramidal layer.

Electrocorticogram (EcoG) recordings and epilepsy induction
EcoG data were acquired at 20 kHz using Epoch Wireless recording system (Ripple, Salt Lake City, USA) and Clampex software (Molecular Devices). Video recording (29 frames/s) was synchronized to data acquisition to enable offline analysis.
For epilepsy induction, animals received intraperitoneal injection of 127 mg/kg of lithium 24 hours before treatment. The following day, animals received i.p. injection of 1 mg/kg of methylscopolamine in order to prevent peripheral effects of pilocarpine. 30 min later, 40 mg/kg of pilocarpine was injected i.p. Both video and EcoG data were acquired for one hour following pilocarpine injection. Status epilepticus was then interrupted by i.p. injection of 5 mg/kg diazepam. Animals were then sacrificed and brain were recovered and processed for visualization of the infected area.              following CNO application (** p<0.01) and repolarized upon Ba 2+ application (** p<0.01).