GABAA receptor dependent synaptic inhibition rapidly tunes KCC2 activity via the Cl--sensitive WNK1 kinase

Abstract

The K⁺-Cl^- co-transporter KCC2 (SLC12A5) tunes the efficacy of GABA_A receptor-mediated transmission by regulating the intraneuronal chloride concentration [Cl^-]_i. KCC2 undergoes activity-dependent regulation in both physiological and pathological conditions. The regulation of KCC2 by synaptic excitation is well documented; however, whether the transporter is regulated by synaptic inhibition is unknown. Here we report a mechanism of KCC2 regulation by GABA_A receptor (GABA_AR)-mediated transmission in mature hippocampal neurons. Enhancing GABA_AR-mediated inhibition confines KCC2 to the plasma membrane, while antagonizing inhibition reduces KCC2 surface expression by increasing the lateral diffusion and endocytosis of the transporter. This mechanism utilizes Cl^- as an intracellular secondary messenger and is dependent on phosphorylation of KCC2 at threonines 906 and 1007 by the Cl^--sensing kinase WNK1. We propose this mechanism contributes to the homeostasis of synaptic inhibition by rapidly adjusting neuronal [Cl^-]_i to GABA_AR activity.

Fig. 1

GABAAR blockade increases KCC2 membrane diffusion. a Examples of KCC2 trajectories showing reduced surface exploration in the presence of muscimol. Scale bars, 0.5 µm. b Median D ±25–75% Interquartile Range IQR of diffusion coefficients of KCC2 in control condition (white) or upon application of muscimol (black) showing no significant effect of muscimol on KCC2 diffusion. n = 416 QDs, 4 cultures, KS test p = 0.139. c, d Time-averaged MSD functions c and median explored area EA ±25–75% IQR d in control (white) vs. muscimol (black) conditions show increased confinement upon muscimol application. n = 838 QDs, 4 cultures, KS test p = 0.039. e Examples of KCC2 trajectories in the presence or absence of gabazine showing increase in QD surface exploration in gabazine-treated condition. Scale bars, 0.5 µm. f Median D of KCC2 in control condition (white) are increased upon gabazine application (black). N = 441 QDs, 5 cultures, KS test p < 0.001. g, h Time-averaged MSD functions g and EA h in control (white) vs. gabazine (black) conditions indicate decreased confinement upon GABAAR blockade. N = 880 QDs, 5 cultures, KS test p < 0.001. i Trajectories (white) overlaid with fluorescent clusters of recombinant homer1c-GFP (green) or gephyrin-mRFP (red) to identify extrasynaptic trajectories (left), trajectories at excitatory (middle) and inhibitory synapses (right). Scale bars, 1 µm. j, k Median D j and EA k of KCC2 are increased upon gabazine application (black) as compared with control condition (white). j extra, n = 129 QDs, KS test p = 0.009, ES, n = 109 QDs, KS test p = 0.004, IS, n = 89 QDs, KS test p = 0.001; 4 cultures. k extra, n = 362 QDs, KS test p < 0.001; ES, n = 212 QDs, KS test p = 0.034; IS, n = 202 QDs, KS test p = 0.015, 4 cultures. l Mean dwell time DT (±s.e.m) of KCC2 at excitatory (ES) and inhibitory (IS) synapses is decreased upon GABAAR blockade with gabazine (black) as compared with control condition (white). ES, Ctrl n = 207 QDs and Gbz n = 218 QDs, MW test p = 0.035; IS, Ctrl n = 162 QDs and Gbz n = 119 QDs, MW test p = 0.047, 4 cultures. b, f, j D in µm²s-1; d, h, k EA in µm²

Fig. 2

Tonic GABAAR but not GABABR activity regulates the lateral diffusion of KCC2. a, b No contribution of tonic GABAAR-mediated inhibition on KCC2 diffusion under basal conditions. Median diffusion coefficients D ±25–75% IQR a and median explored area EA ±25–75% IQR b (for bulk population of QDs) of KCC2 measured in absence (white) or presence (gray) of 50 µM L-655,708 in the absence of exogenous GABA. a n = 320 QDs, 2 cultures, p = 0.658; b, n = 640 QDs, 2 cultures, p = 0.472. c, d Tonic activation of GABAARs by exogenous GABA affects KCC2 diffusion. Application of 2 μM GABA (black) decreased the diffusion and increased the confinement of KCC2 as compared with control condition (white). Addition of L655,708 (pattern) to GABA decreased KCC2 diffusion constraints compared with neurons exposed to GABA only. c n = 271 QDs, 3 cultures; Ctrl vs. GABA p = 0.042; GABA vs. GABA + L-655,708 p = 0.035. d n = 542 QDs, 3 cultures; Ctrl vs. GABA, p < 0.001; GABA vs. GABA + L-655,708 p < 0.001. e, f No effect of GABABR activity on KCC2 diffusion. Median diffusion coefficients e and median explored area f (for bulk population of QDs) of KCC2 measured in absence (white) or presence (black) of baclofen or CGP52432. e Baclofen experiment: n = 278 QDs, 3 cultures, p = 0.864; CGP52432 experiment: n = 279 QDs, 3 cultures, p = 0.425. f baclofen experiment: n = 555 QDs, 3 cultures, p = 0.338; CGP52432 experiment: n = 580 QDs, 3 cultures, p = 0.091. In all graphs, KS test was used for data comparison. a, c, e D in µm²s-1; b, d, f EA in µm²

Fig. 3

Voltage gated Ca²⁺ channels and S940 dephosphorylation do not contribute to gabazine-induced increase in KCC2 diffusion. a, b No effect of gabazine (10 µM) on intracellular calcium level in neurons pre-treated with the Na+ channel blocker tetrodotoxin TTX (1 µM), the ionotropic glutamate receptor antagonist kynurenic acid (1 mM), and the group I/group II mGluR antagonist R,S-MCPG (500 µM). a Pseudocolor images of neurons loaded with Fluo4-AM, before (top) and after (bottom) gabazine treatment. Warmer colors correspond to higher Fluo4-AM fluorescence intensities. Scale bar, 10 µm. b Calcium level in proximal dendrites shown as F/F0 ratio (mean ± s.e.m.) measured as a function of time following gabazine application (black bar). 9 cells; 2 cultures; paired t-test p = 0.691. c, d NMDA (50 µM) increases intra-neuronal calcium level in absence of other drugs. c Pseudocolor images of neurons loaded with Fluo4-AM, before (top) and after (bottom) NMDA treatment. Scale bar, 10 µm. d Calcium level shown as F/F0 ratio as in b. Note the increase in intracellular calcium after NMDA exposure. 19 cells; 2 cultures; paired t test p < 0.001. e, f Median diffusion coefficients D values ±25–75% IQR e and median explored area EA ±25–75% IQR f (for bulk population of QDs) of KCC2 measured in presence of the Ca²⁺ channel blocker Cd²⁺ alone (white) or in presence of gabazine (black). Cd²⁺ did not prevent the gabazine-induced reduction in diffusion constraints of KCC2. e n = 250 QDs, 3 cultures, p = 0.003. f n = 500 QDs, 3 cultures, p < 0.001. g, h Median D g and EA h (for bulk population of QDs) of KCC2-S940D under control (white) or gabazine (black) conditions. Again, S940D substitution did not prevent the gabazine-induced reduction in diffusion constraints of KCC2. g n = 190 QDs, 3 cultures, p = 0.021; H, n = 380 QDs, 3 cultures, p < 0.001. e–h KS test was used for data comparison. e, g D in µm²s-1; f, h EA in µm²

Fig. 4

Changes in intracellular [Cl⁻]_i concentration tune KCC2 diffusion. a Calibration of SuperClomeleon in Neuro2a cells. Ionophore treatment was used to clamp [Cl⁻]_i to 5, 10, 15, 30, 60, and 138 mM. b Images of Cl⁻ dependent changes in the FRET ratio (535/483 nm) in hippocampal neurons expressing SuperClomeleon, before (ctrl) and after 5 or 50 min application of muscimol (Musc), gabazine (Gbz), KCC2 blocker (VU) or extracellular Cl⁻ substitution (0 Cl). The graph shows changes in FRET emission ratio upon treatment relative to control. Muscimol n = 6 cells, p = 0.031 at 5 and 0.063 at 50 min; Gabazine n = 5 cells, p = 0.063 at 5 and 0.008 at 50 min; VU n = 4 cells, p = 0.125 at 5 and 0.015 at 50 min; 0 Cl n = 4 cells, p = 0.047 at 5 and 0.030 at 50 min. Wilcoxon ranked sum test or paired t-test when normality test was passed; 4 cultures. Insets: Ratiometric images of SuperClomeleon in control vs. treatment. Scale bar, 10 µm. c KCC2 trajectories in control vs. VU. Scale bars, 0.5 µm. d Median diffusion coefficient D values ±25–75% IQR (for bulk population of QDs) of KCC2 in control condition (white) or upon application of VU (black). n = 386 QDs, 3 cultures, KS p = 0.228. e, f MSD vs. time functions h and median explored area EA ±25–75% IQR i in control (white) vs. VU (black). N = 712, 3 cultures, KS test p = 0.032. g Matched KCC2 trajectories before (t0), and after 10 s or 1 min of eNpHR (+halo) photostimulation. Scale bars, 0.5 µm. h–j Median D h MSD i and EA j upon 10 s or 1 min of eNpHR photostimulation. h n = 215 QDs, 2 cultures, KS test p = 0.011 and p < 0.001; j, n = 469 QDs, 2 cultures, KS test p < 0.001. k KCC2 trajectories in high vs. low extracellular Cl⁻ concentration. Scale bars, 0.5 µm. l–n Median D l MSD m and EA n showing increased diffusion and reduced confinement of KCC2 upon reduction of Cl⁻ concentration. l n = 408 QDs, 3 cultures, KS test p = 0.001; N, n = 816 QDs, 3 cultures, KS test p < 0.001. d, h, l D in µm²s-1; f, j, n EA in µm²

Fig. 5

WNK1 is involved in KCC2 regulation by GABAAR-mediated transmission. a Left panel, Gel example showing amplification of WNK1 and WNK3 but not of WNK2 and WNK4 cDNAs in DIV 21 hippocampal neurons after 28 cycles of semi-quantitative RT-PCR. a Right panel, Quantitative real time polymerase chain reaction qRT-PCR of DIV 21 cultures illustrating amplification of WNK1 (plain line) followed by WNK3 (cross), WNK4 (circle) and WNK2 (dash) transcripts. b Results of qRT-PCR reactions for WNK1, 2, 3, and 4 from 3-week-old hippocampal cultures (white) and rat adult hippocampal tissue (gray). Expression levels of WNK2, 3 and 4 were normalized to WNK1 expression level. c–e Western blot and quantification (mean ± s.e.m., three independent experiments 1–3) of the activated kinases WNK1 P-S382, SPAK P-S373, and the SPAK homolog OSR1 P-S325 phosphorylation in control (white) and upon gabazine application (black), showing increased phosphorylation of WNK1 (c, d t-test p = 0.011), SPAK (c–e t-test p = 0.017) and OSR1 (c–e t-test p = 0.035) upon GABAAR blockade. f Median diffusion coefficient D values ±25–75% IQR (for bulk population of QDs) of KCC2 show increased KCC2 diffusion upon gabazine application is completely blocked when gabazine is applied in the presence of the SPAK/OSR1 inhibitor closantel. For each condition n = 405 QDs, 4 cultures. Ctrl (white), Gbz (black), closantel (light gray), closantel + Gbz (dark gray), Ctrl vs. Gbz p < 0.001, closantel vs. closantel + Gbz p = 0.55. g Median diffusion coefficients ±25–75% IQR of KCC2 (for bulk population of QDs) in neurons expressing kinase-dead WNK1 (WNK1-KD, light gray) or constitutively-active WNK1 (WNK1-CA, pattern) vs. control plasmid (white). For each condition n = 322 QDs, 5 cultures. Ctrl vs. WNK1-KD p = 0.234, Ctrl vs. WNK1-CA p < 0.001. h WNK1 suppression by shRNA (shWNK1) and overexpression of WNK1-KD abolished the gabazine-mediated increase in KCC2 diffusion in neurons expressing shMock. For each condition n = 322 QDs, 5 cultures. shMock (white), shMock + Gbz (black); shWNK1 (black stripe), shWNK1 + Gbz (white stripe). shMock vs. shMock + Gbz p < 0.001; shWNK1 vs. shWNK1 + Gbz p = 0.812; WNK1-KD vs. WNK1-KD + Gbz p = 0.167. In all f–h graphs, the KS test was used for data comparison. f–h, D in µm²s⁻¹

Fig. 6

GABAAR-dependent regulation of KCC2 membrane dynamics is dependent on KCC2 phospohrylation of T906/T1007. a–c Western blot a and quantification (mean ± s.e.m., three independent experiments 1–3) of KCC2 T906/T1007 b and NKCC1 T203/T207/T212 c phosphorylation in control (white) and gabazine (black) conditions showing increased phosphorylation of KCC2 (t-test T906: p = 0.050; T1007: p = 0.046) and NKCC1 T203/T207/T212 (t-test p = 0.020) upon GABAAR blockade. d Examples of KCC2-T906/T1007, KCC2-T906/T1007A, and T906/T1007E trajectories in control vs. gabazine conditions. Scale bars, 0.5 µm. e Median QD diffusion coefficients D values ±25–75% IQR (for bulk population of QDs) of KCC2-T906/T1007 (white), KCC2-T906/T1007A (gray), and KCC2-T906/T1007E (black stripe) in resting condition show comparable D values of KCC2-T906/T1007 and KCC2-T906/T1007A transporters. In contrast, the phospho-mimetic KCC2-T906/T1007E was faster than KCC2-T906/T1007. T906/T1007 197 QDs, T906/T1007A 238 QDs, T906/T1007E 241 QDs, 3 cultures; T906/T1007 vs. T906/T1007A p = 0.932, T906/T1007 vs. T906/T1007E p < 0.001. f Median D (for bulk population of QDs) of KCC2-T906/T1007, KCC2-T906/T1007A, and KCC2-T906/T1007E in absence or presence of gabazine. Note that gabazine selectively reduced diffusion constraints of KCC2-T906/T1007 but not of KCC2-T906/T1007A and KCC2-T906/T1007E. T906/T1007 (white) n = 197 QDs, T906/T1007 + Gbz (black) n = 197 QDs, T906/T1007A (light gray) n = 238 QDs, T906/T1007A + Gbz (dark gray) n = 238 QDs, T906/T1007E (black stripe) n = 241 QDs, T906/T1007E + Gbz (white stripe) n = 241 QDs, 3 cultures; T906/T1007 vs. T906/T1007 + Gbz p < 0.001, T906/T1007A vs. T906/T1007A + Gbz p = 0.916, T906/T1007E vs. T906/T1007E + Gbz p = 0.648. e, f unless mentioned the KS test was used for all data comparison; d in µm²s⁻¹

Fig. 7

Regulation of KCC2 membrane clustering and stability by GABAAR-mediated inhibition. a Flag surface staining in hippocampal neurons (DIV 21–23) expressing recombinant KCC2–Flag in absence (Ctrl) or presence of gabazine (Gbz), or muscimol (Musc) for 30 min. Scale bars, 10 μm. Note the loss of KCC2 immunoreactivity after 30 min exposure to gabazine but not muscimol. b Quantifications of KCC2 pixel (black) and cluster (gray) intensity in each experimental condition. Note muscimol had no effect while gabazine decreased KCC2 cluster and pixel intensity. Closantel (Clo) treatment and WNK1 shRNA (shWNK1) overexpression suppressed the gabazine-induced reduction in KCC2 clustering as compared with control and shMock treated neurons, respectively. Values were normalized to the corresponding control values. The MW test was used for data comparison. Muscimol experiment: Ctrl n = 62 cells, Musc n = 58 cells, cluster intensity p = 0.823, pixel intensity p = 0.470; 4 cultures. Gabazine and closantel experiment: Ctrl n = 77 cells, Gbz n = 83 cells, cluster intensity p = 0.006, pixel intensity p < 0.001; Closantel n = 65 cells, cluster intensity p = 0.05, pixel intensity p = 0.016; Closantel + gabazine n = 57 cells, cluster intensity p = 0.441, pixel intensity p = 0.448; 4 cultures. ShWNK1 experiment: shMock, n = 48 cells; shMock + gabazine n = 50 cells, cluster intensity p < 0.001, pixel intensity p < 0.001; shWNK1 n = 40 cells, cluster intensity p = 0.38, pixel intensity p = 0.957; shWNK1 + gabazine n = 53 cells, cluster intensity p = 0.249, pixel intensity p = 0.041; 3 cultures. c Biotinylated surface fraction and total protein expression of KCC2 after 30 min of gabazine or muscimol treatments. d No change of total KCC2 protein level between control (white) and gabazine (black; t-test p = 0.126) or muscimol (gray; t-test p = 0.093) conditions (normalized to beta-tubulin III [TUJ1]). N = 4. e Quantification of the ratio of the surface pool of KCC2 over the total pool of KCC2 in control (white), gabazine (black), and muscimol (gray) conditions showing reduction of surface KCC2 after gabazine (t-test p = 0.003) but not muscimol (t-test p = 0.495) treatment (n = 4 experiments)

Fig. 8

Loss of KCC2 upon GABAAR blockade correlates with increased [Cl⁻]_i. a, b Typical gramicidin-perforated patch-clamp recordings of glycine receptor mediated currents induced by short (100 ms) puff of glycine at different membrane potentials in control conditions (a upper trace; b white) or upon 30 min gabazine application (a lower trace; b black). EGlycine was determined as the intercept of the I–V curve with the x-axis. c Gabazine application induced EGlycine depolarization (t-test p = 0.038) indicating increased [Cl⁻]_i. Data represent the mean ± s.e.m. of 13 cells (Ctrl) and 12 cells (Gbz)

Fig. 9

KCC2-dependent spine volume regulation is modulated by GABAAR-mediated inhibition. a Secondary dendrites of eGFP expressing neurons in control conditions or upon application of gabazine or muscimol. Scale bars, 10 µm. b Median values ±25–75% IQR of spine head area from eGFP expressing neurons in control (white), gabazine (black), or muscimol (pattern). Ctrl n = 1301 spines; Gbz n = 1082 spines; Musc n = 1014 spines, 3 cultures; Ctrl vs. Gbz p = 0.005, Ctrl vs. Musc p = 0.978. c Median values ±25–75% IQR of spine head area from shMock or shWNK1 overexpressing neurons in control (white and black stripes) or gabazine (black and white stripes) conditions, respectively. Note the increase in spine head area upon application of gabazine for shMock (p < 0.001) transfected cells and the decrease (p = 0.009) in spine head volume for shWNK1 overexpressing neurons. ShMock n = 260 spines; shMock + Gbz n = 232 spines; shWNK1 n = 95 spines, shWNK1 + Gbz n = 166 spines, 2 cultures. d Same as in a for neurons expressing eGFP and KCC2-Flag or KCC2-Flag mutated on T 906/1007 to A (T906/T1007A) or glutamate (T906/T1007E). Scale bars, 10 µm. e Overexpression of T906/T1007E mutant (gray) but not T906/T1007A mutant (fine stripe) increased spine head volume compared with WT (white). f No change in spine head volume could be observed upon gabazine application in cells expressing KCC2-Flag T906/T1007A (wide stripe). e, f T906/T1007 n = 1301 spines, T906/T1007A n = 1224 spines, T906/T1007E n = 983 spines, T906/T1007A + Gbz n = 1093 spines, 3 cultures; T906/T1007 vs. T906/T1007E p = 0.014, T906/T1007 vs. T906/T1007A p = 0.099, T906/T1007A vs. T906/T1007A + Gbz p = 0.06. b, c, e, f Spine head area in µm². The KS test was used for all data comparison

Fig. 10

GABAAR-dependent regulation of KCC2 in vivo. a, b Activation of the WNK/SPAK signaling pathway in the adult epileptic brain. Western blot a and quantification (b mean ± s.e.m., four independent experiments 1–4) showing seizure induction with the GABAAR antagonist pentylenetetrazole (PTZ, 75 mg/kg) led to phosphorylation/activation of WNK1 S382 (t-test p = 0.003 and p = 0.071) and SPAK S373 (t-test p < 0.001 and p = 0.012), respectively, in the cortex (gray) and hippocampus (black) of adult mice. This was accompanied by a net increase in the phosphorylation of NKCC1 T203/T207/T212 (MW test p = 0.029) and KCC2T1007 (t-test p = 0.021 and p = 0.002) but not KCC2 T906 (MW test p = 0.057 and p = 0.686) in the cortex and hippocampus, respectively