Neuronal Chloride Regulation via KCC2 Is Modulated through a GABAB Receptor Protein Complex

GABAB receptors are G-protein-coupled receptors that mediate inhibitory synaptic actions through a series of downstream target proteins. It is increasingly appreciated that the GABAB receptor forms part of larger signaling complexes, which enable the receptor to mediate multiple different effects within neurons. Here we report that GABAB receptors can physically associate with the potassium-chloride cotransporter protein, KCC2, which sets the driving force for the chloride-permeable ionotropic GABAA receptor in mature neurons. Using biochemical, molecular, and functional studies in rodent hippocampus, we show that activation of GABAB receptors results in a decrease in KCC2 function, which is associated with a reduction in the protein at the cell surface. These findings reveal a novel “crosstalk” between the GABA receptor systems, which can be recruited under conditions of high GABA release and which could be important for the regulation of inhibitory synaptic transmission. SIGNIFICANCE STATEMENT Synaptic inhibition in the brain is mediated by ionotropic GABAA receptors (GABAARs) and metabotropic GABAB receptors (GABABRs). To fully appreciate the function and regulation of these neurotransmitter receptors, we must understand their interactions with other proteins. We describe a novel association between the GABABR and the potassium-chloride cotransporter protein, KCC2. This association is significant because KCC2 sets the intracellular chloride concentration found in mature neurons and thereby establishes the driving force for the chloride-permeable GABAAR. We demonstrate that GABABR activation can regulate KCC2 at the cell surface in a manner that alters intracellular chloride and the reversal potential for the GABAAR. Our data therefore support an additional mechanism by which GABABRs are able to modulate fast synaptic inhibition.


Introduction
GABAergic synaptic inhibition is mediated by two major receptor systems: ionotropic GABA A receptors (GABA A Rs) and metabo-tropic GABA B receptors (GABA B Rs). GABA A Rs rely on transmembrane chloride gradients to generate fast inhibitory synaptic currents (Kaila, 1994;Payne et al., 2003). GABA B Rs, in contrast, generate slower inhibitory actions via the activation of guanine nucleotide-binding protein (G-protein) signaling pathways (Bettler et al., 2004).
It is becoming increasingly clear that to understand the function and regulation of GABA B Rs requires a more complete understanding of the molecular associations that underlie GABA B R complexes in the brain. For instance, recent proteomic approaches have identified auxiliary subunit proteins that modulate the receptor's agonist response and kinetics of G-protein signaling (Schwenk et al., 2010). GABA B R complexes can also include proteins that are the downstream targets following agonist activation of the receptor (Ciruela et al., 2010b;Park et al., 2010) and proteins that are well placed to control the receptor's dimerization or desensitization Pontier et al., 2006). The identification of molecular partners for the GABA B R has also revealed a wider range of functions. These include associations that enable GABA B R subunits to regulate gene transcription (Nehring et al., 2000;White et al., 2000;Vernon et al., 2001) or the intracellular trafficking of other membrane proteins (Boyer et al., 2009). Further diversity in GABA B R function is also likely to relate to the temporal and spatial regulation of the receptor. Recent reports have indicated that the recycling of GABA B Rs at the cell surface is dynamic and can be modulated through receptor activation, composition, phosphorylation, or degradation (González-Maeso et al., 2003;Fairfax et al., 2004;Grampp et al., 2007Grampp et al., , 2008Laffray et al., 2007;Vargas et al., 2008;Wilkins et al., 2008;Hannan et al., 2011).
Here we identify and investigate a novel association between postsynaptic GABA B Rs and the potassium-chloride cotransporter protein, KCC2. KCC2 contributes to the low intracellular chloride concentrations found in mature neurons and thus establishes the conditions for the hyperpolarizing effect of GABA A Rs (Rivera et al., 1999). Furthermore, KCC2 is a locus for modulating the strength of fast synaptic inhibition. Rapid changes in KCC2 function have been shown to be elicited in an activity-dependent fashion and involve different post-translational regulation of the transporter protein, including its phosphorylation state and regulation at the cell surface (Woodin et al., 2003;Rivera et al., 2004;Fiumelli et al., 2005;Lee et al., 2007;Wake et al., 2007;Lee et al., 2010;Chamma et al., 2012;Puskarjov et al., 2012;Medina et al., 2014;Mahadevan and Woodin, 2016).
Using a combination of proteomic, biochemical, and molecular studies, we demonstrate that GABA B Rs and KCC2 can functionally associate with one another at the membrane of neurons. Activation of the GABA B R results in reduced levels of KCC2 at the cell surface, which parallels an increase in intracellular chloride and depolarizing shift in the reversal potential for the GABA A R. Our data support a novel mechanism by which GABA B Rs can modulate KCC2 and thereby fast synaptic inhibition mediated by the ionotropic GABA A R.

Materials and Methods
Mass spectrometry. All experiments using animal tissue were in accordance with regulations from the United Kingdom Home Office Animals (Scientific Procedures) Act. Cortical membranes were prepared by dissecting the cortex from 5 male adult (2 months old) Sprague Dawley rats (Harlan) and homogenizing in 0.32 M sucrose, 50 mM Tris-HCl, pH 7.4 (10 ml/g tissue). The homogenate was centrifuged for 10 min at 690 ϫ g av , 4°C, and the supernatant centrifuged for 20 min at 8700 ϫ g av , 4°C. Each pellet was resuspended in 0.32 M sucrose and layered at the top of a sucrose gradient (0.85 M to 1.0 M to 1.2 M sucrose in 50 mM Tris-HCl, pH 7.4). Gradients were centrifuged for 2 h at 111,000 ϫ g av , 4°C. The membranes were removed, resuspended in 0.32 M sucrose, and centrifuged for 20 min at 19,500 ϫ g av , 4°C. Each pellet was resuspended in 50 ml of cold dH 2 O with protease inhibitors and placed on ice for 30 min. The samples were centrifuged for 20 min at 34,700 ϫ g av , 4°C, and the pellets resuspended in 50 mM Tris-HCl, pH 7.4, before determining the protein concentration and freezing at Ϫ80°C. For each affinity purification (3 in total), 10 mg of the prepared membranes was solubilized in 50 mM Tris-HCl, pH 7.4, containing 1% sodium deoxycholate, protease inhibitors (Boehringer EDTA free), and 10 mM iodoacetamide. The detergent to protein ratio was 5:1. Lysates were centrifuged for 1 h at 66,700 ϫ g av , at 4°C. The supernatant was divided equally and rotated with either 5 g of GABA B R1 antibody or sheep IgG for 6 h at 4°C. A 40 l suspension of Protein G Sepharose Fast Flow beads (GE Healthcare; 1:1 in 50 mM Tris-HCl, pH 7.4) was added and rotated overnight at 4°C. Beads were washed three times with 50 mM Tris-HCl, pH 7.4, ϩ 1% deoxycholate, once with 50 mM Tris-HCl, pH 7.4, and eluted into 25 l Novex 2ϫ reducing sample buffer by heating at 60°C for 15 min. Samples were analyzed by electrophoresis on 4%-12% Bis-Tris NuPAGE Novex gels with MOPs running buffer (Invitrogen). The gel was stained with GelCode Blue Stain Reagent (Pierce), and bands from both experimental and control lanes were excised. Samples were reduced with dithiothreitol, alkylated with iodoacetamide, and digested with trypsin using a MassPREP workstation (Waters). The resulting peptide mixtures were analyzed by liquid chromatography tandem mass spectrometry (MS/ MS) using a CapLC and Q-Tof mass spectrometer (Waters) operating in data-dependent MS/MS mode at the facility at GlaxoSmithKline. Peptides and proteins were identified by automated searching of all MS/MS spectra against a GlaxoSmithKline nonredundant protein database. Candidate proteins associated with the GABA B R: (1) had to appear in three independent isolates, (2) be identified at a position on the SDS-PAGE gel that corresponded to their native molecular weight, (3) have been identified on the basis of two or more peptides on each occasion, and (4) not repeatedly appear in IgG control precipitates.
Preparation of organotypic hippocampal slices. Electrophysiological recordings, live cell imaging, and biochemistry experiments were conducted in organotypic hippocampal brain slices generated from P7 male Wistar rats and cultured for 7-14 DIV before experimentation. Organotypic hippocampal slices were generated as described previously (Stoppini et al., 1991). Briefly, P7 rat brains were extracted and placed in cold (4°C) Geys Balanced Salt Solution (Sigma), supplemented with D-glucose (34.7 mM). The hemispheres were separated, and individual hippocampi were removed and immediately sectioned into 350-m-thick slices on a McIlwain tissue chopper. Slices were rinsed in cold dissection media, placed onto Millicell-CM membranes, and maintained in culture media containing 24.5% v/v EBSS, 49% v/v MEM, 24.5% v/v heat-inactivated horse serum, 0.64% w/v glucose, and 2% v/v B27 (all from Invitrogen; 350 -360 mOsm) at 36°C in a 5% CO 2 humidified incubator. The organotypic hippocampal brain slice enabled us to conduct electrophysiological, imaging, and biochemical experiments in the same preparation. A potential source of variance when investigating chloride homeostasis mechanisms in acutely prepared brain slices has been associated with neuronal damage caused during the slicing procedure (Dzhala et al., 2012;Puskarjov et al., 2012). An advantage of the organotypic hippocampal brain slice is that any neurons that are damaged by the slicing process are lost during the culturing period. Indeed, previous work has shown that the pyramidal neurons in the organotypic hippocampal brain slice have mature and stable chloride homeostasis mechanisms, as evidenced by their hyperpolarizing E GABAA (Ilie et al., 2012;Ellender et al., 2014). This is supported by the current work, which observed that E GABAA is affected by KCC2-blocking drugs, but not by NKCC1-blocking drugs (see below). At the time of electrophysiological recording (P7 ϩ 7-14 DIV), CA3 pyramidal neurons in the organotypic hippocampal slices exhibited a hyperpolarizing E GABAA (Ϫ82.8 Ϯ 1.4 mV) compared with their resting membrane potential (Ϫ71.5 Ϯ 0.9 mV, n ϭ 13; p Ͻ 0.001), and their E GABAA shifted to more depolarized values upon application of 1 mM furosemide (E GABAA in furosemide ϭ Ϫ70.2 Ϯ 2.9 mV; n ϭ 12) or 25 M VU0240551 (E GABAA in VU0240551 ϭ Ϫ75.8 Ϯ 2.9 mV) (Delpire et al., 2009). This is consistent with KCC2 being active in these neurons and contributing to a mature and hyperpolarizing E GABAA . Although many aspects of organotypic hippocampal slice cultures have been shown to resemble the in vivo state (De Simoni et al., 2003), excitatory neurons in this experimental system exhibit increased axonal sprouting, which is likely to underlie the higher levels of synchronous network activity (Dyhrfjeld-Johnsen et al., 2010).
Coimmunoprecipitation. Organotypic hippocampal slices or transfected CHO cells were homogenized in CHAPS buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 5 mM EDTA, 0.5% w/v CHAPS, and protease inhibitors; Roche). Precleared lysates were probed for GABA B R1, KCC2, GFP, or IgG. Protein A/G ϩ agarose was added for 2 h before washing in CHAPS buffer. Agarose beads were eluted in 2ϫ sample buffer at 60°C for 10 min, before loading on to 6% or 8% SDS-PAGE gels. Gels were immunoblotted onto Protran nitrocellulose membranes (Sigma) and probed with indicated primary antibodies overnight at 4°C, before addition of relevant secondary HRP-conjugated antibodies and development with Pierce ECL substrate (Thermo Fisher Scientific).
Biotinylation of cell surface proteins. Rat organotypic hippocampal slices were incubated for 20 min at 28°C-30°C in either control ACSF or ACSF containing 5 M SKF97541 while continuously bubbling with 95% O 2 -5% CO 2 . For biotinylation of both slices and CHO cells, every subsequent step was performed on ice. Samples were incubated for 30 -45 min with 100 M cleavable biotin (EZ-Link Sulfo-NHS-SS-Biotin, Thermo Fisher Scientific), then washed twice with 100 M lysine and lysed with lysis buffer (20 mM Tris, pH 7.5, 50 mM NaCl, 1 mM EDTA, 0.1% w/v SDS, 1% v/v Triton X-100 containing protease inhibitors; Roche). The lysate was centrifuged, and 50 l of the resultant supernatant was removed as the "total" protein lysate sample. Biotinylated proteins were captured by incubation with washed NeutrAvidin Ultralink Resin (Thermo Fisher Scientific) on a rotator overnight at 4°C. The beads were washed 3 ϫ with lysis buffer and the "Surface" sample eluted at 37°C for 30 min in 2ϫ sample buffer. Prepared protein samples were subjected to SDS-PAGE/immunoblotting, as described above. In the CHO cell experiments, fluorescent signals were analyzed using a LI-COR Odyssey scanner. For slice experiments, the ECL signal was captured digitally using a Fluor-S MultiImager (Bio-Rad). Background intensity was subtracted and the optical density for each band quantified through Quantity One version 4.1.0 software (Bio-Rad).
For biotinylation experiments in organotypic hippocampal slices, each sample was comprised of 3 slices from the same animal, maintained on the same Millicell-CM membrane. Every SKF97541-treated sample was processed in parallel with a control sample from the same animal. Between 2 and 8 samples were generated from an individual animal, and each experimental drug manipulation used tissue from between 2 and 6 animals. For each sample, the surface protein was normalized against the total protein, which was run in the adjacent lane. As the ratio of surface/ total was calculated within each sample, this controlled for differences in overall protein levels across samples and variance associated with loading. "Control" and corresponding "SKF97541-treated" samples were always run on the same gel; control values were set to 100% and the SKF97541 treatment expressed as a percentage of control. If the surface/ total ratio for a particular protein was consistently lower for SKF97451treated samples than their corresponding control samples, this would result in a population mean Ͻ100% and would indicate that GABA B R activation caused a decrease in surface levels of the protein.
Immunofluorescence. Organotypic hippocampal slices (P7 ϩ 7-14 DIV) were fixed either in ice-cold methanol (for KCC2 labeling) or in 4% PFA followed by cold methanol (for GABA B R2 labeling). Slices were blocked in PBS containing 0.3% Triton X-100 and 5% normal goat serum. Incubation with primary antibodies (1/1000 dilution for both rabbit anti-GABA B R2 and rabbit anti-KCC2) was performed at 4°C overnight. Slices were washed 4 times with PBS containing 0.3% Triton X-100 and incubated for 4 h at room temperature in the same buffer supplemented with 5% normal goat serum and containing either a 568or 488-coupled anti-rabbit secondary antibody (Invitrogen). Slices were then washed a further 4 times before mounting in 50% glycerol/PBS.
Immunofluorescence was also examined in dissociated hippocampal neurons, where antibody and optical access is better, and where we were able to develop a protocol to quantify coexpression of both proteins within the same cell. Rat dissociated hippocampal cells were prepared at embryonic day 18 (E18) as described previously (Pooler et al., 2009). After 18 -21 DIV cells were fixed and permeabilized in ice-cold methanol, blocked in donkey serum, and sequentially colabeled for GABA B R2 and either KCC2 or ␤-tubulin. Cells were incubated with the rabbit anti-GABA B R2 primary antibody (1/200 dilution), followed by incubation with a donkey anti-rabbit Cy3-monovalent Fab fragment (1:250 dilution). After washing, the cells were further colabeled for KCC2 (1/ 1000) or ␤-tubulin (1/500; both anti-rabbit) followed by incubation with donkey anti-rabbit Alexa-488-conjugated secondary antibody. Appropriate controls were performed to ensure that the Cy3 Fab fragment blocked all available GABA B R sites. Coverslips were washed and mounted with Vectashield mounting medium (Vector Laboratories). Images were collected using a Zeiss Plan-Apochromat 63ϫ, 1.4 NA oil objective, mounted on a Zeiss LSM510 confocal scanning microscope, mounted on an Axiovert 100M inverted microscope (Carl Zeiss). Hippocampal neurons in the dissociated cultures were identified as having a large soma and dendritic spines.
Electrophysiological recordings. Organotypic hippocampal slices were transferred to a recording chamber and continuously superfused with 95% O 2 /5% CO 2 ACSF, heated to 28°C-30°C. These conditions ensured thermal stability and permitted long-term patch-clamp recordings from CA3 pyramidal neurons. The ACSF was composed of the following (in mM): 120 NaCl, 3 KCl, 2 MgCl 2 , 2 CaCl 2 , 1.2 NaH 2 PO 4 , 23 NaHCO 3 , 11 D-glucose, pH 7.3-7.4. With the exception of the synaptic stimulation experiments, the ACSF also contained 1 M TTX (Tocris Bioscience) to eliminate any potential effects at the network level. For perforated patch recordings, the internal solution contained the following (in mM): 135 KCl, 4 Na 2 ATP, 0.3 Na 3 GTP, 2 MgCl 2 , and 10 HEPES, osmolarity 290 mOsm, pH 7.35. Gramicidin (Calbiochem) was added on the day of the experiment to achieve a final concentration of 80 g/ml. Recordings were made with 2-4 M⍀ pipettes via an Axopatch 1D amplifier (Molecular Devices), once perforation had reached a steady access resistance of between 20 and 60 M⍀. To measure E GABAA , cells were maintained at a holding potential of Ϫ60 mV, from which they received voltage steps ranging from Ϫ30 to Ϫ90 mV. Reported membrane potentials were corrected for the voltage drop across the series resistance for each neuron. The liquid junction potential associated with the perforated-patch recordings was small (2.7 mV), and so membrane potential values were not adjusted for this parameter. GABA A R activation was achieved by pressure application of muscimol (10 M, Tocris Bioscience) via a picospritzer (General Valve). To minimize errors associated with access resistance, muscimol-evoked currents were kept small (corresponding to conductances of up to 40 nS) by adjusting the position of the muscimol pipette. Each voltage step lasted for 8000 ms, and cells were returned to the resting holding potential of Ϫ60 mV for 30 s between steps to allow time for the intracellular chloride to reequilibrate (Ehrlich et al., 1999). As a further precaution, the direction in which these voltage steps progressed (i.e., from Ϫ30 to Ϫ90 mV, or from Ϫ90 to Ϫ30 mV) was alternated to avoid any bias in the E GABAA calculations brought about by chloride loading or removal (Akerman and Cline, 2006). Consistent with the fact that transient changes in E GABAA caused by transmembrane fluxes of chloride recover with a time constant of ϳ15 s (Raimondo et al., 2012), our protocol produced reliable estimates of steady-state E GABAA . All drugs were added to the ACSF, with the exception of pertussis toxin (PTX, Sigma) and okadaic acid (Tocris Bioscience), which were added directly to the tissue culture media before experimentation. As required, the following antagonists and blockers were added to the external bathing solution; SKF7541, CGP55845, SCH23390, K252a, D-AP5, kynurenic acid, SR95531, and VU0240551 (all from Tocris Bioscience), furosemide, bumetanide, Gö6976, sodium orthovanodate (Na 3 VO 4 ), and nimodipine, thapsigargin, and monodansylcadaverine (DC) (all from Sigma).
For synaptic stimulation experiments, glutamatergic transmission was blocked by adding 2 mM kynurenic acid to the ACSF, and GABA release at synaptic terminals was evoked by delivering electrical stimuli via a bipolar tungsten stimulating electrode (FHC), placed 50 -100 m from the recorded pyramidal cell, at the border of the stratum pyramidale and stratum radiatum (Scanziani, 2000). To establish the stimulation conditions under which synaptic GABA A R and GABA B R responses are evoked, a series of recordings were first performed in whole-cell mode using a low-chloride internal solution containing the following (in mM): 140 K-gluconate, 2 Na 2 ATP, 3 Na 3 GTP, 2 MgCl 2 , 1 EGTA, and 5 HEPES. To improve detection of GABAergic currents in the whole-cell recordings, cells were clamped at Ϫ50 mV, and GABA A R and GABA B R conductances were calculated by dividing the isolated currents by their driving force (see Fig. 7C).
Synaptic E GABAA was determined from gramicidin perforated patch recordings using a step voltage-protocol from a holding potential of Ϫ60 mV. The holding potential of the cells was stepped at 5 mV increments between Ϫ60 and Ϫ90 mV, during which pure GABA A R currents were elicited using single presynaptic stimuli (see Fig. 7). Again, 30 s was allowed between presynaptic stimuli to allow time for the intracellular chloride to reequilibrate (Ehrlich et al., 1999). In synaptic conditioning experiments, a stimulation protocol was used to strongly activate GABA B Rs (bursts of 6 stimuli at a frequency of 20 Hz, repeated every 5 s for a period of 75 s; see Fig. 7) and the effects upon synaptic E GABAA were measured. During the GABA B R synaptic conditioning protocol, the postsynaptic neuron was held at its E GABAA to avoid transient loading of the cells with chloride (see Fig. 7B). In addition, to allow time for any transient changes to intracellular chloride to fully reequilibrate Ehrlich et al., 1999), the first measurement of synaptic E GABAA following the GABA B R stimulation protocol was made after 5 min. As before, synaptic E GABAA was measured using single presynaptic stimuli to activate pure GABA A R currents.
Wherever possible, electrophysiological recordings were conducted under a "within cell" experimental design. This means that each neuron had E GABAA measurements before ("baseline") and after drug treatment, so that each neuron served as its own control. This "within-cell" experimental design reduces the impact of cross cell variability, means that drug effects can be expressed as a change in E GABAA , and also means that effects can be examined using paired statistical tests. For clarity, we report the mean absolute values of E GABAA and the mean change in E GABAA . To minimize the potential effect of changes in recording conditions that may have taken place over the course of the study, we took the additional step of restricting comparisons to recordings that were performed during similar time periods. For example, over the course of the study, the effect of SKF97541 upon E GABAA was measured in a total of 20 neurons. However, when compared with another experimental group, the SKF97541 data were restricted to recordings performed during a similar time period as the experimental group. Finally, to avoid potential contamination effects across experiments, only one electrophysiological recording was performed per organotypic hippocampal brain slice. This meant that, for each experiment, the number of neurons corresponds to the number of slices. The slices for an individual experiment were generated from between 4 and 10 animals, depending on the particular sample size and complexity of the experiment.
Measurements of intracellular chloride with the Cl-sensor protein. The cyan and yellow fluorescent protein (CFP-YFP) based ratiometric Cl-Sensor protein (Markova et al., 2008) was delivered to CA3 pyramidal neurons in organotypic hippocampal slice cultures by biolistic transfection (Bio-Rad). At 2-3 d after transfection, Cl-Sensor protein expressing neurons were imaged using an FV300 confocal microscope (Olympus), custom-converted for multiphoton imaging, and equipped with a MaiTai-HP Ti:sapphire femtosecond pulsed laser (Newport Spectra-Physics). Images were acquired using Fluoview software (version 5.0, Olympus). Cells were excited at 850 nm and a 510 nm dichroic mirror was used to separate emitted light into CFP and YFP channels, which were filtered at 460 -500 nm and 520 -550 nm, respectively, and detected simultaneously using two externally mounted PMTs (Hamamatsu). Image stacks were flatfield corrected, collapsed along the z-plane, background subtracted, and the YFP/CFP ratio was calculated by dividing the respective images on a pixel-by-pixel basis. The ratio was calibrated to absolute intracellular chloride values using the K ϩ /H ϩ exchanger nigericin and the Cl Ϫ /OH Ϫ exchanger tributyltinchloride (both at 20 M) in a high K ϩ , HEPESbuffered solution at pH 7.35, as described previously (Boyarsky et al., 1988;Kuner and Augustine, 2000).
Data analysis and statistics. Data and statistical analyses were performed using MATLAB R2008b (The MathWorks) and SPSS (IBM). All data are reported as mean Ϯ SE. Statistical comparisons were made using either paired or unpaired Student's t tests, and one-way ANOVAs with post hoc Dunnett (two-sided) corrections. All statistical tests were twotailed, and a p value of Ͻ0.05 was deemed statistically significant.

GABA B Rs form a protein complex with the potassium-chloride transporter KCC2 at the neuronal membrane
We used a combination of coimmunoprecipitation and mass spectrometry to identify functionally important components of GABA B R protein complexes in the brain. An anti-GABA B R1 antibody was used to isolate antibody-protein complexes from membrane preparations generated from freshly dissected adult rat cortex (see Materials and Methods). Analysis of the resulting peptides revealed a series of proteins that have previously been shown to associate with the GABA B R, including G-protein subunits (Bettler et al., 2004), potassium channel tetramerization proteins (Schwenk et al., 2010), NEM-sensitive fusion protein (Pontier et al., 2006), and 14-3-3 signaling proteins (Couve et al., 2001) (Fig. 1A). In addition, the mass spectrometry revealed a novel potential association between the GABA B R and the solute carrier family 12, member 5 protein (SLC12A5), also known as the potassium-chloride cotransporter KCC2 (Fig. 1A). As with all of the associated proteins, KCC2 was present in three independent neuronal membrane isolates, where it was identified from multiple peptides on each occasion and did not appear in IgG control precipitates (Fig. 1B).
KCC2 coimmunoprecipitated robustly with GABA B R1 when protein complexes were isolated from rat organotypic hippocampal brain slices using either a GABA B R1 (Fig. 1B) or a KCC2 antibody (Fig. 1C). KCC2 coimmunoprecipitated with the two splice isoforms of the GABA B R1 subunit: GABA B R1a and GABA B R1b (Fig. 1C). KCC2 appeared as two bands (130 and 270 kDa) on immunoblots ( Fig. 1 B, C), consistent with previous reports that KCC2 can exist as both a monomer and a dimer (Blaesse et al., 2006;. This confirmed that the association between GABA B R and KCC2 is present in organotypic hippocampal rat brain slices as well as acutely dissected rat cortex, and revealed that the GABA B R associates with both monomeric and dimeric forms of KCC2, although it is not clear whether this is a direct interaction or whether other proteins are involved. Consistent with the biochemical evidence, immunofluorescence staining in rat organotypic hippocampal slices (P7 ϩ 7-14 DIV) and rat dissociated hippocampal cultures (E18 ϩ 18 -21 DIV) confirmed that the GABA B R and KCC2 are both found at somatic and dendritic membranes and exhibit overlapping labeling (Fig. 1 D, E). The immunofluorescence protocol in dissociated cultures enabled us to examine coexpression of both proteins within the same cell (see Materials and Methods; Fig.  1E). We therefore quantified the GABA B R and KCC2 staining pattern in the dissociated hippocampal neurons and found that the vast majority exhibited overlapping labeling at the membrane, consistent with colocalization of the proteins (91%; 52 of 57 GABA B R-positive cells).
To further characterize the association between GABA B Rs and KCC2, we used a heterologous CHO cell line that constitutively expresses rat GABA B R1b and GABA B R2 (CHO GABA B R1b/R2) (Pontier et al., 2006). GABA B R1b from this cell line was detected as multiple bands on immunoblots, consistent with differential glycosylation of the GABA B R1b protein in this system ( Fig. 2A). CHO GABA B R1b/R2 was transfected with recombinant versions of rat KCC2 fused to GFP. Coimmunoprecipitation experiments using antibodies against GABA B R1, KCC2 (Fig. 2B), or GFP ( Fig.  2E) confirmed that the association between GABA B Rs and fulllength KCC2 (FL-KCC2) could be reconstituted in this system. KCC2 is predicted to consist of a cytoplasmic amino-terminal domain and a cytoplasmic carboxy-terminal domain, either side of a transmembrane domain that contains 12 transmembrane helices (Fig. 2C) (Payne et al., 1996). We generated GFP fusions of KCC2 that were comprised of only the NTD, only the CTD, or only the TMD. In addition, we generated GFP fusions of KCC2 that lacked either the amino-terminal domain (TMD ϩ CTD) or the carboxy-terminal domain (NTD ϩ TMD). Biotinylation experiments confirmed that each of the fusion proteins containing the transmembrane domain was trafficked, at least in part, to the cell surface in CHO cells (Fig. 2D). This is consistent with previous studies that have shown that KCC cytoplasmic domains (N and C terminal) are not essential for membrane delivery in heterologous cell systems (Casula et al., 2001;. We then performed coimmunoprecipitation experiments to establish the KCC2 region responsible for the association with GABA B Rs. These revealed that GABA B Rs can form a complex with versions of KCC2 that lack both intracellular terminal domains. However, GABA B Rs do not associate with versions of KCC2 that lack the transmembrane domain (Fig. 2E). The association appears spe- Figure 1. GABA B Rs associate with KCC2 at the cell membrane in cortex and hippocampus. A, Affinity purification and mass spectrometry were used to identify proteins associated with the GABA B R in synaptic membrane preparations from adult rat cortex. The table shows the number of unique peptides for each protein, obtained across three independent isolates of the GABA B R1 protein. Where the identified proteins have previously been shown to associate with the GABA B R, references are provided. B, KCC2 coimmunoprecipitates with the GABA B R. Rat hippocampal lysates were immunoprecipitated with an anti-GABA B R1 antibody, and subsequent Western blot analysis for KCC2 revealed two distinct bands at ϳ130 and 270 kDa, which correspond to the monomeric and dimeric forms of KCC2 (top blot, right lane). Probing for GABA B R1 confirmed successful immunoprecipitation of both the GABA B R1a and GABA B R1b isoforms (bottom blot). In contrast, controls using sheep anti-IgG antibody failed to pull down KCC2 or GABA B R1 (middle lanes). C, The GABA B R coimmunoprecipitates with KCC2. Immunoprecipitates from rat hippocampal lysates isolated with an anti-KCC2 antibody were positive for GABA B R1a, GABA B R1b, and KCC2 (top and bottom blots, right lanes). In contrast, control experiments using a rabbit anti-IgG failed to pull down KCC2 or GABA B R1 from the same lysate (middle lanes). D, The GABA B R (left) and KCC2 (right) are both localized at the plasma membrane of pyramidal neurons, as revealed by rabbit polyclonal antibodies against GABA B R2 (left) and KCC2 (right) in separate rat organotypic hippocampal slices (P7 ϩ 7-14 DIV). Scale bars, 10 m. E, Using a sequential double-labeling technique in dissociated neuronal cultures (see Materials and Methods), GABA B R2 (red) and KCC2 (green) were found to be colocalized (yellow) at somatic and dendritic membranes of hippocampal pyramidal neurons. Magnifications of the areas highlighted within the white boxes are provided in the panels below. Control staining (bottom) for GABA B R2 and ␤-tubulin revealed nonoverlapping signals. Scale bars, 20 and 5 m. The GABA B R1 monoclonal antibody 2D7 used for immunoblotting is specific for GABA B R1. In Western blots of total rat brain and rat hippocampus homogenate, two bands corresponding to GABA B R1a and GABA B R1b were detected. Blots of lysates from CHO cells stably expressing either GABA B R1a/R2 or GABA B R1b/R2 revealed multiple bands, consistent with differential glycosylation of GABA B R1 proteins in this system. B, The KCC2-GABA B R association can be reconstituted in a heterologous cell system. CHO cells stably expressing GABA B R1b/R2 were transiently transfected with FL-KCC2-GFP and used for coimmunoprecipitation experiments. Western blots of resulting complexes showed that KCC2 can be coimmunoprecipitated with GABA B R1b (left), and the reciprocal coimmunoprecipitation confirms the association (right). IgG controls were included in each experiment. Sh, Sheep; Rab, rabbit. C, Schematic diagram of KCC2 showing the intracellular NTD, CTD, and TMD with its 12 predicted transmembrane helices. D, GFP fusion proteins of FL-KCC2 and different KCC2 deletion constructs containing the TMD are expressed and transported to the plasma membrane. Biotinylation experiment comparing total (T) and cell surface (S) protein levels in CHO GABA B R1b/R2 cells, transiently transfected with different GFP fusion constructs (left). FL-KCC2 (predicted molecular mass 150 kDa, although the TM region is glycosylated), TMD ϩ CTD (predicted 140 kDa), NTD ϩ TMD (predicted 96 kDa), and TMD (predicted 86 kDa) are all detected on the cell surface, whereas GFP (27 kDa) alone is not. Additional bands detected likely represent alternatively glycosylated, degraded, or aggregated proteins. The blot was reprobed with GABA B R1 antibody to confirm surface expression of the receptor (bottom). E, The transmembrane domain of KCC2 is required for the association with the GABA B R. Coimmunoprecipitation experiments on CHO GABA B R1b/R2 cells transiently expressing KCC2-GFP deletion constructs were performed using anti-GFP as the precipitating antibody. When the resulting complexes were probed for GABA B R1 (top), all KCC2-GFP fusion proteins containing the TMD successfully coimmunoprecipitated GABA B R1b. However, GFP fusion proteins containing only the NTD (predicted 38 kDa) or the CTD (predicted 82 kDa) did not capture GABA B R1b. Under these conditions, and compared with FL-KCC2 (100%), (Figure legend continues.) cific as another transmembrane protein, the transferrin receptor, was not found in the KCC2 isolates (Fig. 2F ). Thus, KCC2 associates with the GABA B R via its transmembrane domain, which is consistent with the idea that KCC2 and GABA B R can form a protein complex at the neuronal membrane.

GABA B R activation affects transmembrane chloride gradients
Signaling interactions across GABA B R protein complexes have been shown to be capable of modulating the activity of both the receptor and its associated proteins (Balasubramanian et al., 2004;Pontier et al., 2006;Ciruela et al., 2010b;Park et al., 2010). Given the evidence that the GABA B R and KCC2 can associate at the membrane, we investigated whether activation of the GABA B R can influence how KCC2 contributes to transmembrane chloride gradients. To assess KCC2 function, intracellular chloride concentration ([Cl Ϫ ] i ) was monitored by calculating the reversal potential of the ionotropic GABA A R (E GABAA ). To avoid disrupting [Cl Ϫ ] i , gramicidin perforated patch-clamp recordings were conducted from CA3 pyramidal neurons in rat organotypic hippocampal slices (P7 ϩ 7-14 DIV). Neurons were clamped at a series of membrane potentials, and GABA A R currents were evoked by delivering brief puffs of the selective GABA A R agonist muscimol (10 M) onto the cell soma. The mean resting membrane potential was Ϫ71.5 Ϯ 0.9 mV, compared with an E GABAA value of Ϫ82.5 Ϯ 1.4 mV (n ϭ 13). Thus, the neurons displayed a mature hyperpolarizing E GABAA profile at the time of recording, consistent with KCC2 expression and function.
To investigate whether agonist activation of GABA B Rs mediates functional changes in [Cl Ϫ ] i , E GABAA was measured before and after the application of the specific GABA B R agonist, SKF97541 (1 M). GABA B R activation was found to result in a depolarizing shift in E GABAA , which was evident 5-10 min following GABA B R activation and persisted for the remainder of the recording (Fig. 3A-D). Across a population of CA3 pyramidal neurons, the mean E GABAA shifted from a baseline value of Ϫ82.5 Ϯ 1.4 mV to Ϫ78.2 Ϯ 1.3 mV following GABA B R activation. This represented a mean change in E GABAA of 4.2 Ϯ 0.7 mV ( p ϭ 0.0009, n ϭ 13, paired t test; Fig. 3D). In control experiments, blocking GABA B Rs with a selective, competitive antagonist (5 M CGP55845) prevented the change in E GABAA in response to SKF97541 ( p ϭ 0.2, n ϭ 6, paired t test). This confirmed that the effects were specific to the GABA B R, and not alternative receptors such as the GABA C R.
Further experiments established that the GABA B R-mediated effect requires associated G-proteins, but is independent of downstream, G-protein-coupled inwardly rectifying potassium (GIRK) channels. First, G-protein signaling via the GABA B R was disrupted by pretreating the organotypic hippocampal slices with the G i / 0 -protein antagonist PTX (5 g/ml for 24 h before recordings). We could confirm that PTX treatment did disrupt GABA B R G-protein-coupled signaling because the SKF97541evoked membrane current that is associated with GIRK channel activity was significantly smaller in PTX-treated neurons (9.4 Ϯ 12.8 pA at a holding potential of Ϫ60 mV, n ϭ 7) than in control neurons (86.0 Ϯ 11.4 pA, n ϭ 13, p ϭ 0.0003, ANOVA followed by post hoc Dunnett's correction; Fig. 3E). Importantly, when we examined the SKF97541-induced effect upon E GABAA , we found that this was significantly inhibited after PTX treatment. The mean E GABAA baseline for PTX-treated cells was Ϫ75.6 Ϯ 3.1 mV and showed little change following SKF97541 treatment when it was Ϫ76.6 Ϯ 3.1 mV. This was significantly different to the effect observed in control cells ( p ϭ 0.034, n ϭ 13 and n ϭ 7, respectively, ANOVA followed by post hoc Dunnett's correction; Fig.  3F ). In contrast, using SCH23390 (10 M) (Kuzhikandathil and Oxford, 2002) to block the downstream GIRK channels directly, did not prevent the SKF97541-induced change in E GABAA . The efficacy of the GIRK channel block by SCH23390 was evident from the significant reduction in the SKF97541-evoked current (35.5 Ϯ 11.3 pA in SCH23390-treated neurons compared with 86.0 Ϯ 11.4 pA in control neurons, n ϭ 8 and 13, respectively, p ϭ 0.01, ANOVA followed by post hoc Dunnett's correction; Fig. 3E). However, under these conditions of GIRK channel block, the mean E GABAA still showed a depolarizing shift from Ϫ81.0 Ϯ 2.2 mV to Ϫ73.6 Ϯ 3.7 mV following SKF97541 application, which was a similar shift to that observed in control cells ( p ϭ 0.24, n ϭ 7, ANOVA followed by post hoc Dunnett's correction; Fig. 3F ). A separate set of experiments revealed that E GABAA was not affected by the activation of postsynaptic adenosine receptors, the most abundant being the metabotropic A1 receptor, which like the GABA B R, is G i / 0 -coupled and targets rectifying potassium channels (Takigawa and Alzheimer, 2002;Ciruela et al., 2010a). The mean E GABAA under baseline conditions was Ϫ84.5 Ϯ 0.9 mV and was not significantly different following activation with adenosine (100 M) at Ϫ83.7 Ϯ 0.9 mV (p ϭ 0.3, n ϭ 13, paired t test). These data support the conclusion that the effect of GABA B R activation upon E GABAA is specific to a GABA B R protein complex but does not involve the activation of downstream potassium channels.
To confirm that the effect of GABA B R activation was via changes in [Cl Ϫ ] i , CA3 pyramidal neurons in organotypic hippocampal slices were transfected with a FRET-based chloride reporter protein called "Cl-Sensor" (Markova et al., 2008;Fig. 3G). Although there was no change in chloride concentration in control neurons expressing the Cl-Sensor protein ( p ϭ 0.67, paired t test, n ϭ 13; Fig. 3H ), activation of GABA B Rs with SKF97541 (5 M for 20 min) resulted in a significant increase in the [Cl Ϫ ] i (p ϭ 0.03, paired t test, n ϭ 7; Fig. 3H). The change in [Cl Ϫ ] i was significantly greater in the SKF97541-treated neurons than in the control neurons imaged over the same time period (p ϭ 0.02, n ϭ 7 and 13, respectively, t test; Fig. 3I). Together, these data show that activation of the GABA B R can lead to a change in intracellular chloride regulation that is consistent with a decrease in KCC2 function.

GABA B R activation can regulate KCC2 at the membrane
To directly test the hypothesis that GABA B R activation modulates KCC2 function at the membrane, we performed a combination of electrophysiological recordings and biotinylation experiments. First, to establish that the GABA B R-mediated shift in E GABAA occurs through a reduction in KCC2 function, cells were exposed to furosemide, which blocks KCC2 activity. CA3 pyramidal neurons exposed to furosemide (1 mM) exhibited a significantly more depolarized resting E GABAA (Ϫ70.2 Ϯ 2.9 mV, n ϭ 12) than untreated control cells (Ϫ83.0 Ϯ 1.9 mV, n ϭ 9, p ϭ 0.002, ANOVA followed by post hoc Dunnett's correction; Fig. 4C). The furosemide-induced shift in E GABAA was evident within 5 min, highlighting that KCC2 functions to continuously maintain the hyperpolarized E GABAA under these conditions. Importantly, ap-4 (Figure legend continued.) the relative amounts of GABA B R detected after isolation with the GFP antibody was 68 Ϯ 16% for NTD ϩ TMD, 36 Ϯ 2% for TMD ϩ CTD, 76 Ϯ 19% for TMD, 3 Ϯ 1% for NTD, 3 Ϯ 2% for CTD, and 3 Ϯ 1% for GFP (n ϭ 2-4 in each case). These quantifications will be influenced by the level of expression of each of the different constructs. Cell lysates expressing the GFP fusion proteins are shown (bottom). Additional bands detected with the GFP antibody are likely to be alternatively glycosylated, degraded, or higher-order aggregates of the expressed fusion proteins. F, Control experiments show that the endogenous 100 kDa transferrin receptor (TfR) is not coimmunoprecipitated with KCC2-GFP proteins in CHO GABA B R cells (top; IgG bands are shown for clarity). Lysates are also shown (bottom). plication of the GABA B R agonist SKF97541 in the presence of furosemide failed to produce any further change in E GABAA (Ϫ70.0 Ϯ 2.3 mV) (Fig. 4 A, D). Compared with control cells, the effect of SKF97541 upon E GABAA was significantly attenuated in furosemide-treated cells (0.2 Ϯ 1.3 mV, p ϭ 0.039, ANOVA followed by post hoc Dunnett's correction; Fig. 4F ).
As furosemide can block multiple cotransporter proteins, this experiment could not exclude a contribution by the sodium-potassium-chloride cotransporter protein, NKCC1, which has also been shown to contribute to [Cl Ϫ ] i regulation in hippocampal pyramidal neurons, particularly during development (Dzhala et al., 2005). We therefore examined the effect of GABA B R activation in the presence of bumetanide, a more selective blocker of NKCC1. The baseline E GABAA after incubation with bumetanide was similar to untreated control cells, indicating that NKCC1 does not make a major contribution to the [Cl Ϫ ] i measured in these neurons (E GABAA ϭ Ϫ82.3 Ϯ 2.3 mV, n ϭ 12, p ϭ 0.97, ANOVA followed by post hoc Dunnett's correction; Fig. 4 B, C). Furthermore, bumetanide treatment did not prevent the depolarizing shift in E GABAA following GABA B R activation. Upon SKF97541 application, the mean E GABAA shifted to a new value of Ϫ77.8 Ϯ 2.7 mV, a change that was indistinguishable to that seen in control cells (4.5 Ϯ 1.1 mV, p ϭ 0.98, ANOVA followed by post hoc Dunnett's correction; Fig. 4 E, F ). GABA B R activation led to an equivalent shift in E GABAA in cells treated with both 10 M bumetanide (baseline E GABAA ϭ Ϫ84.3 Ϯ 3.2 mV, SKF97541-treated E GABAA ϭ Ϫ80.2 Ϯ 4.2 mV, change ϭ 4.1 Ϯ 1.5 mV; p ϭ 0.99, n ϭ 6) or 100 M bumetanide (baseline E GABAA ϭ Ϫ80.2 Ϯ 3.3 mV, SKF97541-treated E GABAA ϭ Ϫ75.4 Ϯ 3.3 mV, change ϭ 4.8 Ϯ 1.7 mV; p ϭ 0.92, n ϭ 6, ANOVA followed by post hoc Dunnett's correction). These data demonstrate that KCC2 is the mediator of the GABA B R-dependent effect upon E GABAA . If this is the case, one prediction is that other manipulations that downregulate KCC2 protein levels should attenuate the Figure 3. GABA B R activation causes a depolarizing shift in E GABAA and increase in intracellular chloride. A, Example GABA A R I-V plot from a gramicidin perforated patch recording of a CA3 pyramidal neuron in a rat organotypic hippocampal slice. Inset, Raw current traces recorded at different holding potentials. GABA A R currents were evoked by local application of muscimol (10 M, arrowhead) to the cell soma. E GABAA was defined as the holding potential at which the GABA A R current had an amplitude of zero. Calibration: 100 pA, 1 s. B, Example I-V plot from the same cell in A recorded after 10 min of bath application of the GABA B R agonist, SKF97541 (1 M). C, GABA B R activation led to a significant depolarizing shift in E GABAA across a population of neurons (n ϭ 13). ***p ϭ 0.0009 (paired t test). Each connected pair of dots corresponds to an individual neuron. Gray horizontal bars represent population means. D, Change in E GABAA plotted as a function of time before and after the onset of SKF97541 application (gray bar; n ϭ 13). E, Disrupting G-protein signaling with PTX or blocking GIRK channels with SCH23390 resulted in a significant reduction in the SKF97541-evoked current compared with control neurons (n ϭ 7, n ϭ 8, and n ϭ 13, respectively). **p Յ 0.01 (ANOVA followed by post hoc Dunnett's correction). F, PTX (n ϭ 7) significantly reduced the shift in E GABAA following GABA B R activation compared with control cells (n ϭ 13). *p ϭ 0.03 (ANOVA followed by post hoc Dunnett's correction). In contrast, blocking GIRK channels did not prevent the shift in E GABAA (n ϭ 7). p ϭ 0.24 (ANOVA followed by post hoc Dunnett's correction). G, Chloride concentration images of a CA3 pyramidal neuron expressing the Cl-Sensor protein and recorded before (left) and after (right) bath application of SKF97541. Scale bar, 10 m. H, Although there was no change in chloride concentration in control neurons 4 expressing the Cl-Sensor protein and exposed to vehicle for 20 min (left; ϩVeh, p ϭ 0.67, paired t test, n ϭ 13), activation of GABA B Rs with SKF97541 (5 M for 20 min) resulted in a significant increase in [Cl Ϫ ] i (right;ϩSKF, *p ϭ 0.03, paired t test, n ϭ 7). I, Activation of GABA B Rs caused a significant increase in [Cl Ϫ ] i after 20 min compared with control cells imaged over the same time period (*p ϭ 0.02, t test). effects of GABA B R-mediated activation upon E GABAA . To test this idea, we maintained organotypic hippocampal brain slices in a zero Mg 2ϩ ACSF for 3 h, which has been shown to cause a robust reduction in KCC2 levels and a depolarizing shift in E GABAA (Puskarjov et al., 2012). Consistent with previous reports, the zero Mg 2ϩ ACSF resulted in a depolarizing shift in E GABAA (to Ϫ64.5 Ϯ 4.9 mV). Importantly, activating GABA B Rs with SKF97541 in slices that had been treated in this manner caused no further change in E GABAA (Ϫ66.6 Ϯ 4.4 mV, n ϭ 7, p ϭ 0.28, t test). Together, these data confirm that KCC2 is the principle mediator of the GABA B R-dependent effect upon [Cl Ϫ ] i .
To investigate whether the GABA B R-mediated decrease in KCC2 function involved regulation of the transporter protein at the cell membrane, we used biotinylation methods to isolate surface KCC2 from rat organotypic hippocampal slices. This approach has been used widely as a way to quantify changes in the plasmalemmal level of chloride transporter proteins in neuronal tissue Thomas-Crussells et al., 2003;Lee et al., 2007;Zhao et al., 2008). Having isolated surface KCC2 protein, we used quantitative Western blot analysis to express the amount of surface KCC2 protein as a ratio of the total KCC2 protein (surface/total KCC2; see Materials and Methods). Experiments revealed that, after 20 min of SKF97541 (5 M) application, there was a reduction in KCC2 levels at the cell surface compared with control slices that were run in parallel but not exposed to the GABA B R agonist (Fig. 5 A, D). Monomeric and dimeric forms of KCC2 were detected at the cell surface, and each showed a significant reduction with SKF97541 exposure, consistent with the hypothesis that GABA B R activation leads to a reduction in membrane bound KCC2 proteins. Monomeric KCC2 at the cell surface was reduced to 80.7 Ϯ 5.1% of control levels ( p ϭ 0.002, n ϭ 14, t test), whereas dimeric KCC2 was reduced to 83.3 Ϯ 7.7% of control levels ( p ϭ 0.048, n ϭ 14, t test; Fig. 5 A, D). The effect was specific to KCC2 as surface levels of the transferrin receptor (94.2 Ϯ 21%; p ϭ 0.79, n ϭ 8, t test) and NKCC1 (92.2 Ϯ 8.9%; p ϭ 0.4, n ϭ 13, t test) were not statistically different between control and SKF97541-treated slices (Fig. 5C,D).
The effects upon surface KCC2 did not appear to involve degradation as the total amount of KCC2 protein was not different between control and SKF97541-treated slices. When examined after the same period of SKF97541 application (5 M for 20 min), the normalized levels of total KCC2 monomer were 104.0 Ϯ 7.1% of control and total KCC2 dimer was 103.7 Ϯ 10.2% of control ( p ϭ 0.58 and p ϭ 0.72, respectively; n ϭ 14, t test). Further, when we repeated these experiments at the later time point of 30 min of SKF97541 application, we again found no evidence for degradation (the normalized levels of total KCC2 monomer were 102.6 Ϯ 9.7% of control and total KCC2 dimer was 114.3 Ϯ 18.5% of control, p ϭ 0.79 and p ϭ 0.46, respectively; n ϭ 12, t test). Interestingly, at this later time point, an effect upon surface KCC2 was also not detectable (surface/total KCC2 monomer was 104.1 Ϯ 8.3% of control and surface/total KCC2 dimer was 106.3 Ϯ 9.9% of control, p ϭ 0.63 and p ϭ 0.54, respectively; n ϭ 12, t test). These data suggest that GABA B R activation does not lead to KCC2 degradation but can rather affect the surface trafficking (endocytosis and recycling) of KCC2. GABA B R dynamics at the cell surface can be affected upon receptor activation (Laffray et al., 2007;Grampp et al., 2008;Wilkins et al., 2008), suggesting that changes to surface trafficking of KCC2 could be associated with changes to surface GABA B Rs. Indeed, our electrophysiological recordings revealed that SKF97541-evoked currents tended to decrease in the continued presence of the agonist (a decrease of 10.1 Ϯ 3.7%, from 88.5 Ϯ 12.8 pA to 79.0 Ϯ 11.1 pA, with 10 min of SKF97541 exposure; n ϭ 13, p ϭ 0.012, paired t test), indicating that there may be an agonist-dependent change in GABA B R signaling at the cell surface. To investigate this biochemically, we used our quantitative Western blot methods to examine GABA B R behavior after exposure to SKF97541 and found that, at the same time point that surface KCC2 is reduced ( Fig. 5 A, D), levels of surface GABA B Rs were also reduced. SKF97541 treatment (5 M for 20 min) reduced GABA B R1 surface expression in the organotypic hippocampal slices to 79.4 Ϯ 8.7% of control ( p ϭ 0.046, n ϭ 9, t test; Fig.  5 B, D). Meanwhile, coimmunoprecipitation experiments revealed that the amount of KCC2 pulled down in GABA B R1 complexes was not significantly different between SKF97541treated and matched control slices ( p ϭ 0.44, n ϭ 3, t test; Fig.  5E). To investigate whether the internalization of KCC2 via GABA B Rs could be reconstituted in the heterologous CHO cell system, we applied SKF97541 to CHO GABA B R1b/R2 transiently expressing FL-KCC2. In this system, there was no significant effect upon surface levels of either GABA B R1 (105.7 Ϯ 3.4% of control; p ϭ 0.12, n ϭ 13, t test) or KCC2 (98.8 Ϯ 6.1% of control; p ϭ 0.85, n ϭ 13, t test), suggesting that the mechanism may be sensitive to expression levels, post-translational modifications, or that intermediary proteins are involved in regulating surface expression in neurons. Figure 5. GABA B R activation leads to a reduction in KCC2 at the membrane surface. A, GABA B R activation leads to a reduction in the fraction of both monomeric and dimeric forms of KCC2 at the membrane surface. Rat organotypic hippocampal slices (P7 ϩ 7-14 DIV) were exposed to normal ACSF (Control) or ACSF containing SKF97541 for 20 min (SKF, 5 M), and then biotinylated to label proteins at the cell surface. Cell homogenates of total protein (T) and streptavidin-purified cell surface proteins (S) were probed on Western blots with anti-KCC2 antibodies (top). Lack of ␤-tubulin staining in surface samples confirmed that the cell membranes remained intact (bottom). B, SKF97541 treatment also led to a concomitant reduction in surface levels of GABA B R1 relative to controls. C, Neither surface-bound NKCC1 nor the surface levels of the transferrin receptor changed following GABA B R activation. D, Summary plot of the effect of GABA B R activation upon proteins at the membrane surface after 20 min 5 M SKF97541 treatment. Surface expression was quantified as the ratio of surface to total protein, normalized to control slices that were examined in parallel. SKF97541 treatment resulted in a significant reduction in the surface ratio for KCC2 (monomer, **p ϭ 0.002; dimer, *p ϭ 0.048; all, **p ϭ 0.003; n ϭ 14, t test) and GABA B R1 (all, *p ϭ 0.046, n ϭ 9, t test). No effect was observed in the surface/total ratio for NKCC1 ( p ϭ 0.4, n ϭ 13, t test) or the transferrin receptor (TfR; p ϭ 0.79, n ϭ 8, t test). E, The association between KCC2 and GABA B R remains following SKF97541 treatment. Rat organotypic hippocampal slices were solubilized and subjected to coimmunoprecipitation analysis using GABA B R1 as the precipitating antibody. The amount of KCC2 in GABA B R1 complexes (top), normalized by the amount of GABA B R1 protein precipitated (bottom), did not change following SKF97541 treatment ( p ϭ 0.44, n ϭ 3, t test).

GABA B R regulation of KCC2 involves clathrin-mediated endocytosis
Both GABA B Rs and KCC2 undergo endocytosis via the clathrin-mediated endocytotic pathway (Grampp et al., 2007;Laffray et al., 2007;Vargas et al., 2008;Zhao et al., 2008). Indeed, clathrin-mediated endocytosis contributes to the constitutive membrane recycling of GABA B Rs, which can be accelerated by receptor activation (Grampp et al., 2007(Grampp et al., , 2008Laffray et al., 2007). One possibility therefore is that the clathrin-mediated endocytotic pathway is important for the change in surface KCC2 that results from GABA B R activation. To test this, we used a selective blocker of the clathrin-mediated endocytotic pathway, dansylcadaverine (DC). Importantly, CA3 pyramidal neurons pretreated with DC (50 M) failed to show a change in E GABAA following GABA B R activation ( Fig. 6 A, B). E GABAA shifted from a mean baseline of Ϫ82.6 Ϯ 3.4 mV to Ϫ82.1 Ϯ 3.3 mV following SKF97541 treatment, a shift of just 0.46 Ϯ 0.6 mV (n ϭ 8), which was significantly smaller than the change observed in control cells exposed to SKF97541 (n ϭ 11, p ϭ 0.03, ANOVA followed by post hoc Dunnett's correction; Fig.  6E). DC-treated neurons also failed to show a decrease in the amplitude of SKF97541evoked currents in the continued presence of the agonist (a decrease of Ϫ0.9 Ϯ 3.9%, from 36.3 Ϯ 8.3 pA to 35.5 Ϯ 7.8 pA, with 10 min of SKF97541 exposure; n ϭ 8, p ϭ 0.36, paired t test), suggesting no net change in GABA B R signaling at the cell surface. Finally, consistent with these electrophysiological recordings, the reduction in levels of surface KCC2 following GABA B R activation (5 M SKF97541 for 20 min) was reduced when clathrinmediated endocytosis was blocked with DC (50 M; n ϭ 7; p ϭ 0.13; Fig. 6 F, H ). Disrupting the clathrin-mediated endocytotic pathway therefore occludes the GABA B R-mediated change in surface KCC2.
KCC2 function has also been reported to be modulated by Ca 2ϩ -dependent kinases, phosphatases, and proteases (Fiumelli et al., 2005;Lee et al., 2007Lee et al., , 2010Wake et al., 2007;Xu et al., 2008;Puskarjov et al., 2012). To test whether the GABA B R-dependent modulation of KCC2 involves intracellular Ca 2ϩ signaling, hippocampal slices were treated with a combination of Ca 2ϩ channel blockers and intracellular Ca 2ϩ store blockers (20 M nimodipine, 100 M D-APV and 2 M thapsigargin). Under these conditions, GABA B R activation still resulted in a positive shift in E GABAA that Figure 6. The GABA B R-effect upon KCC2 requires clathrin-dependent endocytosis. A, Blocking clathrin-dependent endocytosis with DC prevents the depolarizing shift in E GABAA following GABA B R activation. Example I-V plots are shown for a CA3 pyramidal cell in a rat organotypic hippocampal slice treated with DC (50 M). E GABAA at baseline (black data) is similar to that recorded after SKF97541 treatment (gray data). B, The SKF97541-induced change in E GABAA in a population of CA3 pyramidal neurons treated with DC (black bar) and plotted as a function of the time of SKF97541 application (gray bar; n ϭ 8). C, Blocking Ca 2ϩ signaling via a combination of nimodipine (20 M), thapsigargin (2 M), and D-APV (100 M) did not prevent the GABA B R-mediated shift in E GABAA . Same conventions as in A. D, The SKF97541-induced change in E GABAA in a population of CA3 pyramidal neurons treated with Ca 2ϩ channel blockers (black bar), plotted as a function of the time of SKF97541 application (gray bar; n ϭ 9). E, Summary plot of the effect of GABA B R activation in the presence of different inhibitors. Blocking clathrin-dependent endocytosis significantly reduced the SKF97541-dependent shift in E GABAA (n ϭ 8) observed in control cells (n ϭ 11). *p ϭ 0.03 (ANOVA followed by post hoc Dunnett's correction). In contrast, treating cells with a combination of Ca 2ϩ channel blockers (n ϭ 9) had no effect on the SKF97541-dependent shift in E GABAA . p ϭ 0.99 (ANOVA followed by post hoc Dunnett's correction). Preincubation with the selective protein kinase C inhibitor, Gö6976 (1 M, n ϭ 9, p ϭ 0.99); the general kinase blocker, K252a (100 nM, n ϭ 6, p ϭ 0.94); the tyrosine phosphatase inhibitor, Na 3 VO 4 (1 mM, n ϭ 6, p ϭ 0.99); or the protein phosphatase 1 and 2 inhibitor, okadaic acid (1 M, n ϭ 6, p ϭ 0.99) did not prevent a significant shift in E GABAA following GABA B R activation (all ANOVA followed by post hoc Dunnett's correction). F, Surface KCC2 levels are not altered when GABA B Rs are activated in the presence of an inhibitor of clathrin-dependent endocytosis. Rat organotypic hippocampal slices were pretreated with DC (50 M) and exposed to SKF97541 (DC ϩ SKF). Slices pretreated with DC but not exposed to SKF97541 (DC) were used as controls and run in parallel (see Materials and Methods). G, Slices treated with a combination of Ca 2ϩ blockers still showed reduced surface KCC2 levels following GABA B R activation. H, Summary plot of the ratio of surface-to-total KCC2 protein, normalized to control values. Blocking clathrindependent endocytosis prevented the reduction in surface KCC2 following GABA B R activation (n ϭ 7, p ϭ 0.13, t test), whereas blocking Ca 2ϩ signaling (n ϭ 8) did not prevent the effect of GABA B R activation upon surface KCC2. *p ϭ 0.011 (t test).
was indistinguishable from the shift in control slices (baseline E GABAA ϭ Ϫ78.8 Ϯ 3.5 mV, SKF97541 treatment E GABAA ϭ 74.9 Ϯ 4.3 mV, change ϭ 3.9 Ϯ 0.9 mV, p ϭ 0.99, n ϭ 9, ANOVA followed by post hoc Dunnett's correction; Fig. 6C-E). Blocking Ca 2ϩ channels also failed to prevent the SKF97541-mediated (5 M for 20 min) decrease in surface KCC2 measured by biotinylation (77.9 Ϯ 6.4% p ϭ 0.011, n ϭ 8, t test; Fig. 6G,H ). Consistent with these observations, the positive shift in E GABAA was not prevented by inhibitors of calcium-dependent kinases. Pretreatment with Gö6796 (1 M), a selective inhibitor of calcium-dependent protein kinase C, did not prevent the GABA B R-dependent shift in E GABAA (baseline E GABAA ϭ Ϫ85.0 Ϯ 3.0 mV, SKF97541 treatment E GABAA ϭ Ϫ80.7 Ϯ 2.3 mV, change ϭ 4.3 Ϯ 1.1 mV, p ϭ 0.99, n ϭ 9, ANOVA followed by post hoc Dunnett's correction; Fig. 6E). K252a (100 nM), which inhibits protein kinase A and tyrosine kinases, also failed to block the GABA B R-dependent shift (baseline E GABAA ϭ Ϫ83.3 Ϯ 3.7 mV, SKF97541 treatment E GABAA ϭ Ϫ78.2 Ϯ 3.4 mV, change ϭ 5.0 Ϯ 1.0 mV, p ϭ 0.94, n ϭ 6, ANOVA followed by post hoc Dunnett's correction; Fig. 6E). Similarly, treatment with the tyrosine phosphatase inhibitor sodium orthovanadate (Na 3 VO 4 ; 1 mM) did not prevent a positive shift in E GABAA upon GABA B R activation (baseline E GABAA ϭ Ϫ71.5 Ϯ 3.4 mV, SKF97541 treatment E GABAA ϭ Ϫ67.0 Ϯ 3.7 mV, change ϭ 4.5 Ϯ 1.3 mV, p ϭ 0.99, n ϭ 6, ANOVA followed by post hoc Dunnett's correction; Fig. 6E). Treating cells with the phosphatase 1 and 2A inhibitor okadaic acid (1 M), also failed to block the SKF97541mediated shift in E GABAA observed in control slices (baseline E GABAA ϭ Ϫ88.1 Ϯ 0.7 mV, SKF97541 treatment E GABAA ϭ Ϫ84.9 Ϯ 1.1 mV, change ϭ 3.2 Ϯ 1.2 mV, p ϭ 0.99, n ϭ 6, ANOVA followed by post hoc Dunnett's correction; Fig. 6E). Together, these data demonstrate that activation of the GABA B R leads to a decrease in the surface expression of KCC2, in a . GABA B R responses were not evoked by single stimuli but were evident for the multiple-stimuli condition. In the absence of these receptor blockers (right), the flux of chloride through GABA A Rs could be minimized by clamping the postsynaptic neuron close to its E GABAA . Calibration: 100 pA, 500 ms. C, The amplitude of the postsynaptic GABA B R response is sensitive to presynaptic stimulus frequency. Whereas GABA A R conductances (gGABA A ) were detected across the range of stimulus frequencies, GABA B R-mediated conductances (gGABA B ) were largest for high-frequency stimuli of ϳ20 Hz and were minimal at lower frequencies (n ϭ 9). D, Resting E GABAA values measured by muscimol activation of the GABA A R (n ϭ 25) and by synaptic activation of the GABA A R (n ϭ 22). Synaptic E GABAA exhibited a greater range of values and had a mean value of Ϫ76.7 Ϯ 2.1 mV, compared with Ϫ81.1 Ϯ 1.1 mV for the muscimol-evoked recordings ( p ϭ 0.06, t test). E, Example GABA A R I-V plots for a CA3 pyramidal neuron before (black data) and after (gray data) delivering a conditioning protocol designed to strongly activate postsynaptic GABA B Rs (90 stimuli delivered as 15 bursts of 6 stimuli at 20 Hz, at 5 s intervals). Insets, Raw traces. Calibration: 50 pA, 1 s. F, Change in E GABAA in a population of CA3 pyramidal neurons (n ϭ 6) following delivery of the GABA B R 4 conditioning protocol (vertical arrow). G, CA3 pyramidal neurons that underwent the GABA B R synaptic conditioning protocol (n ϭ 6) showed a significantly larger positive shift in E GABAA than neurons that experienced a control stimulation protocol (90 stimuli delivered at 1 Hz) designed to generate minimal GABA B R activation (n ϭ 6, *p ϭ 0.017, ANOVA followed by post hoc Dunnett's correction). The change in E GABAA induced by the GABA B R synaptic conditioning protocol was also prevented by blocking GABA B Rs with the selective antagonist CGP55845 (n ϭ 5, *p ϭ 0.022) or by blocking KCC2 activity with VU0240551 (25 M; n ϭ 7, ***p ϭ 0.001. manner that is independent of calcium-dependent kinase and phosphatase activity, but is dependent upon clathrin-mediated endocytosis.

Synaptically driven GABA B R activity affects intracellular chloride regulation
To investigate whether this mechanism could be recruited under physiological conditions, we examined whether the GABA B Rmediated effect upon KCC2 occurs at inhibitory synaptic connections. Presynaptic GABAergic interneurons in organotypic hippocampal slices were stimulated via a bipolar electrode placed at the border of the stratum pyramidale and stratum radiatum, 50 -100 m from the recorded pyramidal cell (Fig. 7A). This enabled us to evoke monosynaptic GABA A R responses and to measure synaptic E GABAA . Baseline synaptic E GABAA was similar to muscimol-evoked E GABAA , with a mean value of Ϫ76.7 Ϯ 2.1 mV (n ϭ 22; Fig. 7D). We next examined whether the GABA B Rmediated effect upon E GABAA could be elicited via synaptic activation of GABA B Rs. GABA B Rs are located predominantly extrasynaptically in hippocampal pyramidal cells and are thought to be activated under conditions of strong GABA release, such as occur during periods of high-frequency presynaptic firing (Scanziani, 2000). Consistent with this, a single presynaptic stimulus generated a pure GABA A R response in CA3 pyramidal neurons, which was entirely blocked by SR95531 (10 M; Fig. 7B). In contrast, high-frequency trains of stimuli (e.g., 6 stimuli at 20 Hz) produced a postsynaptic response that was comprised of a large GABA A R conductance and a smaller GABA B R-mediated conductance that could be blocked by CGP55845 (5 M; Fig. 7B). By varying presynaptic stimulation conditions, it was observed that the optimal presynaptic frequency for activating a GABA B R response was close to 20 Hz (Fig. 7C).
Having established the stimulation parameters for isolating the GABA A R response and for evoking robust GABA B R responses, we asked whether synaptically driven GABA B R activation could induce an activity-dependent shift in E GABAA . Using gramicidin perforated patch recordings, baseline E GABAA was first determined by using single presynaptic stimuli to elicit a postsynaptic GABA A R response at different holding potentials (Fig. 7E). A synaptic stimulation protocol was then administered, which had been shown to elicit strong GABA B R activation and consisted of bursts of 6 stimuli at a frequency of 20 Hz, repeated every 5 s for a period of 75 s (GABA B R synaptic conditioning protocol; see Materials and Methods). To avoid loading the cells with chloride during these stimulation trains, the holding potential of the recorded cell was clamped at E GABAA , so that there was minimum flux of chloride through the GABA A R (Fig. 7B). After the GABA B R conditioning protocol, synaptic E GABAA was then remeasured as before using single presynaptic stimuli. These experiments revealed that the GABA B R synaptic conditioning protocol caused a robust depolarizing shift in E GABAA (Fig. 7 E, F ). Across a population of cells, the mean E GABAA shifted from a baseline value of Ϫ73.3 Ϯ 3.4 mV to Ϫ67.2 Ϯ 4.6 mV when recorded 15 min after synaptic GABA B R stimulation, which represented a change in E GABAA of 6.1 Ϯ 1.7 mV ( p ϭ 0.014, n ϭ 6, paired t test; Fig. 7G). A temporal analysis of the data showed that the shift in E GABAA was evident 10 min following synaptic stimulation of the GABA B R (E GABAA ϭ Ϫ69.4 Ϯ 4.1 mV, change ϭ 4.0 Ϯ 1.3 mV, p ϭ 0.027, n ϭ 6, paired t test) and was still detected at 30 min after stimulation, the longest population data point that we were able to record (E GABAA ϭ Ϫ64.5 Ϯ 4.2 mV, change ϭ 8.9 Ϯ 1.4 mV, p ϭ 0.002, n ϭ 6, paired t test; Fig. 7G).
To establish that this effect was dependent upon GABA B R activation, we first confirmed that a control stimulation protocol that generated minimal GABA B R activation (90 stimuli delivered at 1 Hz) did not elicit a change in E GABAA . The baseline E GABAA was Ϫ77.1 Ϯ 4.1 mV; and after delivering the control stimulation protocol, E GABAA was Ϫ75.9 Ϯ 3.0 mV, which was a significantly smaller change in E GABAA (1.2 Ϯ 1.4 mV) than observed after the GABA B R conditioning protocol ( p ϭ 0.017, n ϭ 6, ANOVA followed by post hoc Dunnett's correction; Fig. 7G). Then we confirmed that blocking GABA B Rs with a competitive antagonist (5 M CGP55845) was able to significantly attenuate the shift in E GABAA caused by the GABA B R synaptic conditioning protocol. Indeed, in these experiments, the baseline E GABAA was Ϫ79.3 Ϯ 1.2 mV; and after delivering the GABA B R synaptic conditioning protocol, the E GABAA was Ϫ78.2 Ϯ 1.7 mV. This was a change of 1.1 Ϯ 0.5 mV, which did not represent a significant depolarizing shift in E GABAA (1.1 Ϯ 0.5 mV, n ϭ 5; t test; p ϭ 0.11) and was significantly smaller than the E GABAA change observed without the GABA B R antagonist ( p ϭ 0.022, n ϭ 5, ANOVA followed by post hoc Dunnett's correction; Fig. 7G).
Finally, blocking KCC2 with the selective antagonist VU0240551 (25 M) (Delpire et al., 2009;Ivakine et al., 2013) also reduced any change in E GABAA following GABA B R stimulation with the GABA B R conditioning protocol (baseline E GABAA ϭ Ϫ72.1 Ϯ 3.6 mV, poststimulation E GABAA ϭ Ϫ72.6 Ϯ 2.6 mV, change ϭ Ϫ0.5 Ϯ 1.1 mV, p ϭ 0.001, n ϭ 7, ANOVA followed by post hoc Dunnett's correction; Fig. 7G). These experiments demonstrate that the GABA B R-mediated effect upon E GABAA via KCC2 is not only elicited by exogenous agonist activation of the GABA B R but can also be elicited by synaptically evoked GABA release.

Discussion
By forming signaling complexes through specific interactions with other proteins, G-protein-coupled receptors convert extracellular signals into diverse neuronal responses. In the case of GABA B Rs, this includes G-proteins that are required for their "classic" signaling, but also interactions with auxiliary proteins that modulate the kinetics of receptor signaling (Schwenk et al., 2010), desensitization (Pontier et al., 2006, subunit dimerization , and regulate the localization of the receptor or other proteins within cells (White et al., 2000;Boyer et al., 2009). Here we have identified a novel association between the GABA B R and the potassium-chloride cotransporter KCC2. This association was discovered in an unbiased screen for proteins present within GABA B R complexes at the neuronal membrane, was confirmed by biochemical experiments in hippocampal brain slices and heterologous cells, and was shown to be mediated via the transmembrane region of KCC2. Agonist activation of the GABA B R elicits signaling events at the neuronal membrane via G-protein-coupled complexes. We observed that GABA B R activation led to a rapid and sustained change in the ionic driving force for the chloride-permeable GABA A R, consistent with a decrease in KCC2 function. Electrophysiological recordings and biotinylation assays confirmed that the effects were mediated via KCC2 and were associated with a change in the trafficking of KCC2 protein at the cell surface. A similar downregulation in KCC2 function could also be elicited by a synaptic conditioning protocol designed to strongly activate GABA B Rs. And while other signaling mechanisms may have been activated under our experimental conditions, the principle change in E GABAA was unlikely to be mediated by alternative GABA receptors, such as the GABA C R, or by other signaling systems (Mahadevan and Woodin, 2016), because the effect of the GABA B R agonist and synaptically released GABA were both blocked by a selective GABA B R antagonist. These results are consistent with evidence that GABA B R activation modulates proteins with whom the receptor is physically associated (Ciruela et al., 2010b;Park et al., 2010).
In our recordings from CA3 pyramidal neurons, activation of the GABA B R by agonist or by synaptically released GABA resulted in an ϳ5 mV positive shift in E GABAA , which is similar in amplitude to the shifts in E GABAA following other activitydependent changes to chloride transporter proteins (Woodin et al., 2003;Wang et al., 2006;Xu et al., 2008;Ormond and Woodin, 2009). Shifts in E GABAA were evident within ϳ10 min following GABA B R activation, which is also consistent with previous evidence that E GABAA can be rapidly modulated within minutes (Woodin et al., 2003;Fiumelli et al., 2005;Wang et al., 2006;Balena and Woodin, 2008;Xu et al., 2008). Our longest recordings were unable to capture the reversal of the effects on E GABAA and showed that they were evident for at least 30 min, which is again similar to previous studies that have examined the activitydependent regulation of KCC2 function (Woodin et al., 2003;Fiumelli et al., 2005;Kitamura et al., 2008;Lee et al., 2011;Puskarjov et al., 2012;Zhou et al., 2012). Assuming E GABAA reflects E Cl Ϫ and that extracellular chloride remains constant, a 5 mV shift would equate to an increase in intracellular chloride of ϳ1.2 mM (from 5.4 to 6.6 mM, according to the Nernst equation). Changes in E GABAA over a narrow range (Ͻ5 mV) can have dramatic effects upon whether GABAergic inputs have an inhibitory or facilitating effect (Morita et al., 2006;Jedlicka et al., 2011) and E GABAA changes of the same magnitude can cause significant changes in the degree of NMDA receptor activation and action potential firing frequency (Akerman and Cline, 2006;Saraga et al., 2008), which can be further influenced by the frequency and location of GABAergic inputs (Prescott et al., 2006;Jean-Xavier et al., 2007).
The GABA B R-mediated effect upon KCC2 appears to be distinct from previously described, activity-dependent mechanisms that regulate KCC2. Post-translational regulation of the transporter has been linked to calcium signaling events and associated enzymatic modifications. KCC2 function is associated with its phosphorylation state (Woodin et al., 2003;Fiumelli et al., 2005;Lee et al., 2007;Wake et al., 2007), and the transporter has been reported to turnover rapidly and as a function of the phosphorylation of specific sites within the C-terminal (Rivera et al., 2004;Lee et al., 2010). Recent work has revealed that the total pool of KCC2 is much more stable, but that degradation can be triggered by intracellular calcium, which activates calcium-dependent proteases that cleave the C-terminal of KCC2 (Puskarjov et al., 2012). In contrast to these mechanisms, the GABA B R-mediated effect upon KCC2 was not prevented by blocking calcium signaling processes, it was not affected by blockers of kinases and phosphatases implicated in the regulation of KCC2, and the total levels of KCC2 were not altered, suggesting that degradation pathways are not involved.
The stable physical association we observed between the GABA B R and KCC2, plus evidence that GABA B Rs can exhibit dynamic behavior at the membrane, offers a potential mechanism by which GABA B R activation could influence the surface stability and/or trafficking of the transporter protein. Previous work has provided differing results on the membrane dynamics of GABA B Rs. Some studies have reported that the receptor is stable at the cell surface, regardless of whether it is activated or not (Fairfax et al., 2004;Grampp et al., 2007). Other studies have provided evidence that GABA B Rs are mobile, being rapidly and constitutively internalized on a timescale of minutes via clathrindependent pathways, and in a manner that can be modulated by activation of the receptor (Laffray et al., 2007; Grampp et al., 2008;Wilkins et al., 2008). Our experiments in rat organotypic hippocampal slices revealed that GABA B R activation can result in a decrease in the surface expression of both GABA B R and KCC2 proteins. Such a reduction in surface GABA B R following receptor activation is consistent with previous observations in slice cultures (Laffray et al., 2007) but contrasts with studies in dissociated neuronal cultures (Fairfax et al., 2004;Vargas et al., 2008), suggesting that the experimental system (Vargas et al., 2008), or factors such as the dimerization state of the GABA B R (Laffray et al., 2007;Hannan et al., 2011), may be important. The effects we observed appeared to affect only a subset of the proteins (Ͻ25% decrease in both surface proteins) and were evident over a similar, but not identical, timescale to the downregulation in KCC2 function that we measured electrophysiologically. These differences in timescales of effect may reflect the sensitivities of the methods but could also indicate functional changes to KCC2 that result from being recycled to the membrane, perhaps due to changes in membrane domain, cellular location, and/or molecular interactions (Hartmann et al., 2009;. Internalized GABA B Rs are associated with the clathrin-binding adaptor protein-2 complex (Grampp et al., 2007), and disrupting clathrin-mediated endocytosis prevents internalization and recycling of GABA B Rs (Grampp et al., 2007;Laffray et al., 2007;Vargas et al., 2008). Similarly, KCC2 has been shown to bind to adaptor protein-2 in the brain and to undergo fast clathrin-mediated endocytosis (Zhao et al., 2008). Importantly, we found that blocking clathrin-mediated endocytosis prevented GABA B Rs from downregulating KCC2 function and expression at the neuronal membrane. Together, these data support a model in which active GABA B Rs can modulate the surface stability of KCC2 via a mechanism that involves clathrin-mediated endocytosis and which impacts the transporter's contribution to transmembrane chloride levels. It is worth noting that our data do not demonstrate a direct interaction between KCC2 and GABA B R; therefore, the potential for additional proteins to mediate the functional association in neurons should also be considered.
The fact that we observed a 20% reduction in surface KCC2 and a smaller GABA B R-mediated shift in E GABAA than was produced by furosemide is consistent with the idea that different pools of KCC2 exist, which differ in terms of their localization, protein associations, and/or stability in the membrane. For instance, recent work has revealed that a pool of KCC2 is not localized at GABAergic synapses but rather at glutamatergic postsynaptic structures, where it functionally associates with kainate receptors and has been implicated in regulating glutamatergic transmission (Gauvain et al., 2011;Mahadevan et al., 2014;Chevy et al., 2015). Our experiments did not distinguish between protein complexes located in different subcellular compartments, such as the soma or dendrites. Future experiments could therefore explore whether the KCC2-GABA B R association varies as a function of cellular location or the membrane lipid environment (Hartmann et al., 2009;). It will also be interesting to examine the longer-term consequences of manipulating the KCC2-GABA B R association, where the use of transgenic mouse lines is likely to be informative (Schuler et al., 2001;Vigot et al., 2006).
In conclusion, GABA B Rs are able to associate in a protein complex with the potassium-chloride cotransporter KCC2. Activation of the GABA B R can result in a decrease in KCC2 function, which requires the clathrin-mediated endocytosis pathway, regulates the transporter protein at the cell surface, and alters the driving force for chloride-permeable GABA A Rs. These findings reveal a novel "crosstalk" between the GABA receptor systems,