Lateral Diffusion of NKCC1 Contributes to Chloride Homeostasis in Neurons and Is Rapidly Regulated by the WNK Signaling Pathway

An upregulation of the Na+-K+-2Cl− cotransporter NKCC1, the main chloride importer in mature neurons, can lead to depolarizing/excitatory responses mediated by GABA type A receptors (GABAARs) and, thus, to hyperactivity. Understanding the regulatory mechanisms of NKCC1 would help prevent intra-neuronal chloride accumulation that occurs in pathologies with defective inhibition. The cell mechanisms regulating NKCC1 are poorly understood. Here, we report in mature hippocampal neurons that GABAergic activity controls the membrane diffusion and clustering of NKCC1 via the chloride-sensitive WNK lysine deficient protein kinase 1 (WNK1) and the downstream Ste20 Pro-line Asparagine Rich Kinase (SPAK) kinase that directly phosphorylates NKCC1 on key threonine residues. At rest, this signaling pathway has little effect on intracellular Cl− concentration, but it participates in the elevation of intraneuronal Cl− concentration in hyperactivity conditions associated with an up-regulation of NKCC1. The fact that the main chloride exporter, the K+-Cl− cotransporter KCC2, is also regulated in mature neurons by the WNK1 pathway indicates that this pathway will be a target of choice in the pathology.


Introduction
Upon GABA activation, GABA type A receptors (GABA A Rs) activate a selective chloride/bicarbonate conductance in neurons. The direction of chloride (Cl − ) flux through the channel depends on transmembrane Cl − gradients. Therefore, Cl − homeostasis determines the efficacy of GABAergic transmission. Drug-resistant epilepsies are often associated with defective Cl − transport [1]. It is, therefore, crucial to find new mechanisms regulating neuronal Cl − transport that could eventually lead to the development of treatments for drug-resistant epilepsies and diseases with decreased inhibition, such as neuropathies and neuropsychiatric disorders [2].
The increase in [Cl − ] i and the consequent impact on the GABA reversal potential (E GABA ) in various models of epilepsy is often attributed to a malfunction of the neuronal K + -Cl − cotransporter KCC2, which exports Cl − from neurons [3]. In addition, increased expression and/or activity of the Na + -K + -Cl − cotransporter NKCC1, which imports chloride into neurons, also contributes to increased [Cl − ] i and depolarization of E GABA . This has been observed in the subiculum of temporal lobe epilepsy (TLE) patients [4,5] as well as in several in vivo and in vitro epilepsy models, such as those induced by a glioma [6] or traumatic brain injury [7] models and the kainic acid and pilocarpine epilepsy models in tissue slices [8][9][10]. Targeting NKCC1 in epilepsy with the inhibitor bumetanide is protective: it abolished glioma-induced seizures in rats [6] and reduced the frequency of interictal-like activities [4], the duration of ictal activities [10], and the sprouting of mossy fibers [9].

QD-Based Single-Particle Tracking
Neurons were imaged at 33 • C with an inverted Olympus IX71 microscope, a 60X objective (Numerical Aperture NA 1.42; Olympus, Rungis, France), and a Mercury lamp (X-Cite 120Q, Lumen Dynamics, Mississauga, ON, Canada). The diffusion of NKCC1 along axons was studied on neurons co-transfected with NKCC1 and eGFP plasmids. eGFP labeling allowed for distinguishing dendrites from axons. Indeed, the eGFP staining uniformly fills the soma, dendrites, and axons of the neurons, allowing seeing the overall morphology of the neuron. Mature neurons have very thin and long axons that are easily distinguishable from much thicker and shorter dendrites. Moreover, since the transfection efficiency is very low, it is easy to see the origin of the axon of a given neuron. We, therefore, imaged the axons of isolated neurons in their first hundred micrometers. The diffusion of NKCC1 at synapses vs. at the distance of synapses was followed in neurons transfected with NKCC1 constructs together with GPHN.FingR-eGFP and homer1c-DsRed plasmids. Individual GPHN.FingR-eGFP and homer1c-DsRed images were acquired with an ImagEM Electron-multiplying CCDs (EMCCD) camera and MetaView software (Meta Imaging Series 7.8). QD real-time movies of 1200 frames were acquired with a 30 ms integration time.
QD tracking and trajectory reconstruction was performed with homemade software (Matlab R2020A; The Mathworks, Natick, MA, USA) as described in [26]. Trajectories were considered synaptic when overlapping with the synaptic mask of GPHN.FingR-eGFP or homer1c-DsRed clusters, or extrasynaptic for spots four pixels (760 nm) away. The inclusion area was increased relatively to previous studies on KCC2 [13,18] due to the low numbers of NKCC1 trajectories recorded with a 380 nm distance. Expanding the radius did not actually change results for perisynaptic NKCC1 lateral diffusion, suggesting that its clusters are located further away from synapses than KCC2 ones. One to two dendritic and Cells 2023, 12, 464 5 of 24 axonal regions were analyzed per cell. The QD crossing trajectories were removed from the analysis.
The mean square displacement (MSD) vs. time curves were determined for each trajectory by using the equation: In this equation, τ = acquisition time, N= total number of frames, i = positive integers, and n= time increment. Diffusion coefficients (D) were obtained by fitting the first four points without the origin of the MSD vs. time curves with MSD(nτ) = 4Dnτ + σ, with σ being the spot localization accuracy. The pointing accuracy of the QD is~20-30 nm. The explored area corresponds to the MSD value of the trajectory at two different time intervals of 0.42 and 0.45 s [27]. The dwell time at synapses was defined as the duration of detection of QDs at synapses divided by the number of exits [26]. Dwell times lower than five frames were discarded from the analysis.

Fluorescence Image Acquisition and Analysis
Image acquisition was performed using a 100× objective (1.40 NA) on a Leica (Nussloch, Germany) DM6000 upright microscope with a 12-bit cooled CCD camera (Micromax, Roper Scientific) using MetaMorph software Series 7.8 (Roper Scientific). To assess NKCC1-HA clusters, the exposure time was determined on the brightest experimental condition in order to be non-saturating and was fixed for all cells and conditions to be analyzed. Quantification was performed using a MetaMorph routine (Roper Scientific). For the dendritic intensity and clustering analysis, a region of interest (ROI) was traced around a selected dendrite, and the average pixel intensity in the ROI was measured. For the clustering analysis, images were filtered using the flattened background (kernel size, 3 × 3 × 2) function on Metamorph to enhance cluster outlines, and an intensity threshold Cells 2023, 12, 464 6 of 24 defined by the user was set to identify the clusters and avoid their coalescence. Clusters were outlined, and the corresponding regions were transferred onto the raw images to determine NKCC1-HA cluster number, surface, and fluorescence intensity. To estimate the membrane fraction of NKCC1, the mean pixel intensity of Venus and of Cy3-tagged membrane NKCC1-HA were computed from a dendritic ROI, and the surface/total ratio of intensity was calculated. The area of the dendritic region analyzed was used to calculate the number of clusters per surface area. We analyzed~10 cells per experimental condition and per culture.

STORM Microscopy
STORM was performed on fixed samples on an inverted N-STORM Nikon Eclipse Ti microscope (Nikon, Lisses, France) equipped with a 100× objective (NA 1.49) and an Andor iXon Ultra EMCCD camera, and using specific lasers for STORM of Alexa 647. Image acquisition consisted of the accumulation of 30,000 frames with a 50 ms frame rate. A Nikon perfect focus system was used to maintain the z position during the whole acquisition. Single-molecules were localized, and 2D images were reconstructed as described [28] by fitting the PSF of fluorophores to a 2D Gaussian distribution. The position of fluorophores was corrected by the relative movement of the cluster by calculating the center of mass of the cluster throughout the acquisition using a partial reconstruction of 2000 frames with a sliding window [28]. Rendered images were obtained by superimposing the coordinates of single molecule detections. To correct multiple detections obtained from the same molecule of Alexa 647, detections occurring in the vicinity of space (2σ) and time (15 s) were identified as belonging to the same molecule.
The number of NKCC1 clusters, their size, and the density of detections in the pixels of a cluster were measured on 2D images reconstructed through cluster segmentation based on detection densities. The minimal thresholds to determine clusters were 1% intensity, 0.1 per nm 2 minimum detection density, and 10 detections. The resulting binary image was analyzed with the function "regionprops" of Matlab to extract the surface area of each cluster identified by this function. The molecular density corresponds to the number of detections in the pixels (STORM pixel size = 20 nm) that belong to a cluster divided by the cluster area.

Statistics
N corresponds to the number of QDs for SPT experiments, clusters for STORM, and cells for immunocytochemsitry and chloride imaging. The sample size was determined from published work, pilot experiments, and our own expertise. Mostly all data were used. For SuperClomeleon imaging, SPT or immunocytochemsitry, suffering cells with blobs and/or fragmented neurites were discarded from the analysis. Data representation was performed with boxplots or cumulative frequency plots. The statistical test to compare the two groups was either the Welch t-test when the normality assumption was met (Q-Q plots and cumulative frequency fit); otherwise, the Mann-Whitney test was performed to assess the presence of dominance between the two distributions. For variables following a log-normal distribution, such as those obtained from SPT and STORM assays, we applied the log(.) function after dividing by the control group's median. For super-resolution experiments, as an important variability could be observed between different cells in the same coverslip, a balanced random selection of clusters across neurons, conditions, and cultures was performed, then variables from each culture were divided by the median of the control group. Results from different cultures were pooled, and log(.) was applied, then the Mann-Whitney U value was computed. The process was repeated 1000 times, and the p-value was determined from the U distribution using the basic definition of the p-value. For SPT analysis, note that each QD is associated with 3 EA values; thus, the sample size is three times greater. Statistical analysis was performed with R (3.6.1, R Core Team, 2019, Vienna, Austria) using the following package (ggplot2, matrixStats). Differences were significant for p-values less than 5% (* p < 0.05; ** p < 0.01; *** p < 0.001; NS, not significant).

Opposite Effects of GABAergic Activity on NKCC1 Mobility in the Axon vs. the Dendrite
We questioned whether inhibitory GABAergic transmission influences NKCC1 lateral diffusion in mature hippocampal cultures at 21-23 DIV using QD-SPT. We showed by Western blot in a previous study that mature  hippocampal cultures express endogenous NKCC1 [18]. The expression level of NKCC1 in our cultures does not differ from that in hippocampal tissue, suggesting that this expression is not an artifact of the culture [18]. Here, we determined the impact of a pharmacological activation or blockade of GABA A Rs in the presence of tetrodotoxin (TTX), kynurenate (KYN), and mCPG (MCPG) to block action potentials and glutamatergic activity and compared to the TTX + KYN + MCPG "control" condition. In the absence of antibodies targeting a specific extracellular epitope of NKCC1 at the surface of living neurons, we transfected hippocampal cultures with recombinant HA-tagged NKCC1 constructs at 14 DIV. Then, we stained them at 21-24 DIV with an anti-HA antibody coupled to QDs and imaged them (see Materials and Methods). First, we reported that in spontaneous activity conditions, NKCC1 diffuses along the axon (Supplementary Figure S1A) and in the somato-dendritic compartment ( Figure 1A). Neurons were acutely exposed to the GABA A R agonist muscimol (10 µM) or competitive antagonist gabazine (10 µM), two drugs that were shown using electrophysiological recordings to increase or block GABA A R-mediated inhibition in mature hippocampal neurons [18]. We observed that, when exposed to gabazine or muscimol, QDs explored more the surface of the axon compared to QDs in control conditions ( Figure S1A). The slope of the mean square displacement (MSD) as a function of time was increased for trajectories recorded in gabazine and muscimol conditions compared to controls (Supplementary Figure S1B), indicating reduced confinement of NKCC1 in the axon. This was associated with an increase in the diffusion coefficient (Supplementary Figure S1C) and the explored area (Supplementary Figure S1D). Therefore, NKCC1 confinement is reduced in the axon under conditions of GABAergic activity blockade.
Moreover, we found that GABAergic signaling oppositely regulated the mobility of NKCC1 in the axon vs. the dendrite. In contrast to what we found in the axon, we showed that QDs explored smaller areas of the dendritic membrane following exposure to gabazine or muscimol compared to the control ( Figure 1A). Analysis performed on the whole population (extrasynaptic + synaptic) of dendritic trajectories revealed that the MSD function displayed a less steep slope for trajectories recorded in the presence of muscimol or gabazine as compared with control ( Figure 1B), indicative of increased confinement upon activation or blockade of the GABA A R. The median diffusion coefficient and explored area values of dendritic NKCC1 were also reduced upon muscimol and gabazine application ( Figure 1C,D, respectively). Thus, the lateral diffusion of NKCC1 on the dendrite is regulated by inhibitory GABAergic transmission: the transporters are slowed down and confined in response to a change in GABAergic activity.
We then analyzed the effects of gabazine and muscimol on the diffusion of NKCC1 at extrasynaptic and synaptic sites. Muscimol and gabazine reduced the transporter mobility and surface exploration of individual QDs ( Figure 1E). Quantitative analysis on populations of QDs revealed an impact of the treatments on diffusion coefficient and explored the area for extrasynaptic QDs and for QDs at both excitatory and inhibitory synapses ( Figure 1F,G). The effect was greater on the diffusion coefficient than on the explored area for NKCC1 trajectories located at the periphery of glutamatergic synapses and vice versa for trajectories near inhibitory synapses ( Figure 1F,G). Thus, NKCC1 exhibits increased diffusion constraints at extrasynaptic sites and at the periphery of synapses upon GABA A R activation or blockade.  In all graphs, *, p < 5.0 × 10 −2 ; **, p < 1.0 × 10 −2 ; ***, p < 1.0 × 10 −3 .

NKCC1 Is Targeted to Endocytic Zones Where They Are Stored upon GABAergic Activity Changes
We have shown (unpublished work) that NKCC1, unlike KCC2, is confined within endocytic zones without necessarily being internalized. This suggests that NKCC1 can be stored in endocytic zones. This pool of NKCC1 would constitute a reserve of membrane transporters, which would be rapidly released in the plasma membrane under specific conditions. We observed that acute exposure to muscimol or gabazine increased the confinement of NKCC1 in endocytic zones, as observed in neurons transfected with clathrin-YFP for individual trajectories ( Figure 1H) or for hundreds of molecules ( Figure 1I,J). Moreover, blocking clathrin-mediated endocytosis with an inhibitory peptide prevented the slow down and increased confinement of NKCC1 upon gabazine or muscimol treatment ( Figure 1K,L). We, therefore, concluded that the treatment of neurons with muscimol or gabazine increased the confinement of NKCC1 transporters in endocytic zones in the dendrites.
To investigate whether the increased confinement of the transporter to endocytic zones induced by GABA A R agonists and antagonists is accompanied by an increase in its internalization, we analyzed the surface pool of NKCC1 (measured as the ratio of the mean fluorescence intensity of the surface/surface + intracellular NKCC1) (Figure 2A,B). The ratio remained unchanged ( Figure 2B) after exposure to muscimol or gabazine for 30 min, indicating that changes in GABA A R activity do not affect the membrane stability of NKCC1. This is consistent with the regulation of KCC2 diffusion by GABAergic inhibition, which regulates the transporter clustering and thus function without requiring transporter internalization [18]. We examined whether changes in GABA A R-dependent inhibition could alter NKCC1 clustering. Using conventional epifluorescence, we reported that muscimol significantly reduced by 1.28-fold the density of NKCC1 clusters at the surface of neurons transfected with NKCC1-HA ( Figure 2C). Furthermore, a 30 min exposure to muscimol reduced by 1.1-fold the size of NKCC1 clusters ( Figure 2D), as compared with untreated cells. In contrast, muscimol did not affect the fluorescence intensity of the clusters ( Figure 2E). These results indicate that the increased confinement of NKCC1 in endocytic zones induced upon muscimol treatment correlates with a reduction in its membrane clustering. Unlike muscimol, gabazine did not noticeably alter the density of NKCC1 clusters ( Figure 2C). However, it significantly reduced the size ( Figure 2D) and intensity ( Figure 2E) of these clusters, suggesting transporter loss within clusters.
Since NKCC1 cluster size is at the limit of the resolution of a standard epifluorescence microscope, we further analyzed the effect of the treatments on NKCC1 clustering using super-resolution STORM. NKCC1 forms round-shaped clusters along the dendrites ( Figure 2F). We report that neuronal exposure to gabazine or muscimol altered the nanoscopic organization of NKCC1 ( Figure 2F). The muscimol treatment significantly decreased by 1.16-fold, the average size of NKCC1 nanoclusters ( Figure 2G). This effect was not accompanied by a decreased number of molecules per cluster ( Figure 2H). In fact, the density of particles detected per cluster was significantly increased by 1.43-fold after muscimol exposure ( Figure 2I), indicating molecular compaction. This effect, coupled with the loss of NKCC1 clusters observed with standard epifluorescence and the increased confinement of the transporter in endocytic zones, is consistent with a muscimol-induced escape of NKCC1 transporters from membrane clusters followed by their recruitment and storage in endocytic zones. clustering. Unlike muscimol, gabazine did not noticeably alter the density of NKCC1 clusters ( Figure 2C). However, it significantly reduced the size ( Figure 2D) and intensity (Figure 2E) of these clusters, suggesting transporter loss within clusters.  STORM microscopy revealed that gabazine treatment significantly decreased the size of NKCC1 clusters (by 0.62-fold, Figure 2G), accompanied by a 1.5-fold reduction in the number of molecules per cluster ( Figure 2H), in agreement with the notion that transporters escaped clusters. This effect was not associated with a change in the density of molecules per cluster ( Figure 2I), reporting no change in the compaction of molecules within the cluster. Although the effects of muscimol and gabazine differ on the nanoscale organization of NKCC1, both treatments lead to the escape of transporters from clusters, which are then rapidly captured in the endocytic zones where they are stored.

Intracellular Chloride Levels Tune NKCC1 Surface Diffusion and Clustering
We then investigated whether variations in [Cl − ] i could explain the effects of gabazine and muscimol treatment on NKCC1 diffusion, as shown for KCC2 [18]. We lowered [Cl − ] i by replacing Cl − with methane sulfonate in the imaging medium and tested its effect on NKCC1 diffusion. We showed that in similar 21 DIV-old hippocampal neurons, lowering the extracellular chloride in the medium by replacing it with methane sulfonate significantly increases (+25%) the FRET ratio of the SuperClomeleon signal as expected for a decrease in [Cl − ] i [18]. The decrease in [Cl − ] i increased NKCC1 surface exploration for individual trajectories located at the periphery of synapses and at a distance ( Figure 3A). Quantification revealed that this treatment had no effect on the diffusion coefficients of dendritic NKCC1 for either extrasynaptic or synaptic trajectories ( Figure 3B). On the other hand, lowering [Cl − ] i decreased the confinement of extrasynaptic transporters ( Figure 3C), while the confinement of synaptic transporters was unchanged ( Figure 3C). We then asked if the reduced confinement observed at extrasynaptic sites concerned transporters stored in endocytic zones. We found that the diffusion coefficient and surface exploration of NKCC1 were significantly increased for transporters located at a distance of clathrin-coated pits ( Figure 3D-F) while the mobility of transporters in endocytic zones was not modified upon lowering [Cl − ] i ( Figure 3E,F). Therefore, reducing [Cl − ] i does not increase the confinement of NKCC1 in endocytic zones. Altogether, our results provide evidence that lowering intracellular chloride levels removes diffusion constraints onto NKCC1, which moves faster in the membrane, probably by being relieved from endocytic zones.
We then determined whether this relief in diffusion constraints of the transporter was associated with its membrane redistribution. Quantification of the membrane pool of NKCC1 revealed that lowering [Cl − ] i increased NKCC1 immunoreactivity on the dendrites ( Figure 3G). This was correlated with a 1.25-fold increase in the surface amount of NKCC1 ( Figure 3H). This was also accompanied by an increase in its clustering. The treatment did not alter the number of clusters detected at the cell surface ( Figure 3I). However, lowering [Cl − ] i increased by 1.12-fold the size of NKCC1 clusters ( Figure 3J) and by 1.4-fold their fluorescence intensity ( Figure 3K). These results indicate a chloride-dependent regulation of NKCC1 diffusion-capture, consistent with homeostatic regulation of the transporter.

The WNK Signaling Pathway Regulates NKCC1
The chloride-sensitive WNK signaling pathway regulates NKCC1 activity [21]. We assessed the role of this pathway in NKCC1 clustering using overexpression of constitutively active (WNK-CA) or kinase-dead (WNK-KD) WNK1 [24] and using WNK1 (WNK463) inhibitor. Overexpression of WNK-CA has no effect on the area explored by individual QDs (Figure 4A) or by hundreds of QDs ( Figure 4B,C). Similarly, overexpression of WNK-CA did not alter the level of NKCC1 detected on the surface ( Figure 4D,E) or the number of NKCC1 clusters ( Figure 4F), or the size and intensity of these clusters ( Figure 4G,H). Based on these results, we concluded that, under basal activity conditions, activation of the WNK1 pathway does not affect the diffusion, amount, and distribution of NKCC1 at the dendritic surface in mature hippocampal neurons. We then determined whether this relief in diffusion constraints of the transporter was associated with its membrane redistribution. Quantification of the membrane pool of NKCC1 revealed that lowering [Cl − ]i increased NKCC1 immunoreactivity on the dendrites ( Figure 3G). This was correlated with a 1.25-fold increase in the surface amount of (WNK463) inhibitor. Overexpression of WNK-CA has no effect on the area explored by individual QDs (Figure 4A) or by hundreds of QDs ( Figure 4B,C). Similarly, overexpression of WNK-CA did not alter the level of NKCC1 detected on the surface ( Figure 4D,E) or the number of NKCC1 clusters ( Figure 4F), or the size and intensity of these clusters ( Figure 4G,H). Based on these results, we concluded that, under basal activity conditions, activation of the WNK1 pathway does not affect the diffusion, amount, and distribution of NKCC1 at the dendritic surface in mature hippocampal neurons.  Conversely, we studied the effects of inhibition of the WNK1 signaling pathway on NKCC1 surface expression and clustering following overexpression of WNK-KD or after acute exposure to a pan-WNK antagonist (WNK-463). An acute blockade for 30 min of WNK1 with WNK463 had no effect on the membrane stability of NKCC1 ( Figure 4I,J), while blocking WNK1 activity for 7 DIV by overexpressing WNK-KD significantly reduced the membrane stability of NKCC1 ( Figure 4I,J). On the other hand, WNK inhibition using genetic or pharmacological approaches significantly altered the clustering of NKCC1 by decreasing, respectively, to 2-fold and 1.6-fold the number of NKCC1 clusters ( Figure 4K). This was not followed by a reduction in cluster size upon WNK463 treatment or WNK-KD overexpression ( Figure 4L). However, WNK-KD overexpression decreased to 2-fold the intensity of NKCC1 clusters ( Figure 4M). Therefore, inhibiting the WNK1 signaling pathway in basal activity conditions reduces NKCC1 membrane stability and clustering.
We then studied the contribution of the WNK1 effectors SPAK and OSR1 in the control of NKCC1 membrane diffusion, stability, and clustering using the SPAK/OSR1 inhibitor closantel [29]. Acute exposure of neurons to closantel rapidly reduced the dendritic exploration of individual QDs ( Figure 5A). This was accompanied by a 1.14-fold reduction in NKCC1 diffusion coefficients ( Figure 5B) and by a 1.28-fold decrease in its explored area ( Figure 5C), revealing increased NKCC1 diffusion constraints as compared with control. This effect on diffusion was, however, not accompanied by a change in the surface detection of NKCC1 ( Figure 5D,E), nor by a change in its clustering as determined by standard epifluorescence on the number of NKCC1 clusters ( Figure 5F), as well as on their size and intensity ( Figure 5G,H). However, the analysis of NKCC1 clusters using super-resolution imaging ( Figure 5I) revealed that closantel exposure was reduced by 1.3-fold the cluster size ( Figure 5J). This effect was not linked to a change in the number of clusters per dendritic length ( Figure 5K). However, closantel increased by 1.75-fold the density of molecules per cluster ( Figure 5L) as compared with untreated cells, indicative of molecular compaction. Therefore, the closantel-induced confinement of NKCC1 is accompanied by a rapid alteration in the nanoscale organization of the transporter. These results report that the WNK1/SPAK/OSR1 pathway regulates the membrane dynamics, stability, and clustering of NKCC1 in mature neurons. The fact that WNK1 activation (by overexpressing WNK-CA) has no effect on NKCC1 membrane dynamics, expression, and clustering suggests that this pathway is active in mature neurons and regulates NKCC1 diffusion-capture.

The WNK Signaling Pathway Targets Key Threonine Residues on NKCC1
WNK kinases promote the phosphorylation of NKCC1 T203/T207/T212 [24], which in turn results in NKCC1 activation [21]. To test the involvement of NKCC1-T203/207/212/217/ 230 phosphorylation status in the control of NKCC1 diffusion, we expressed NKCC1 mutations of T203/T207/T212 or T203/T207/T212/T217/T230 into alanine (TA3 and TA5, respectively) that mimic dephosphorylated states. The mobility and exploration of individual NKCC1 T203/207/212/217/230A were decreased relative to WT, especially for extrasynaptic QDs ( Figure 6A). This was reflected by a 1.2-fold lower speed (Figure 6B), and a 1.48-fold increased confinement ( Figure 6C) of QDs in the extrasynaptic membrane without changing the diffusion coefficient or the surface area explored at the inhibitory and excitatory synapses ( Figure 6B,C). Therefore, the dephosphorylation of NKCC1 on key threonine residues confines the transporter in the extrasynaptic membrane. In agreement with a regulation of NKCC1 diffusion-capture by the WNK1 signaling pathway, these results indicate that a proportion of NKCC1 is phosphorylated on T203/207/212 in mature neurons. This differs from the KCC2 transporter, for which regulation by the WNK1 signaling was only observed when GABA A R activity was challenged [18].  membrane without changing the diffusion coefficient or the surface area explored at the inhibitory and excitatory synapses ( Figure 6B,C). Therefore, the dephosphorylation of NKCC1 on key threonine residues confines the transporter in the extrasynaptic membrane. In agreement with a regulation of NKCC1 diffusion-capture by the WNK1 signaling pathway, these results indicate that a proportion of NKCC1 is phosphorylated on T203/207/212 in mature neurons. This differs from the KCC2 transporter, for which regulation by the WNK1 signaling was only observed when GABAAR activity was challenged [18].  We previously showed, in similar cultures, that an acute application of muscimol increases [Cl − ] i [18]. This is corroborated by the inhibition of WNK1 and dephosphorylation of NKCC1 [18]. Knowing that the phosphorylation of NKCC1 by WNK1 regulates its activity in non-neuronal cells [30], we wanted to know if the membrane stability and clustering of the mutated transporter were altered compared to that of the WT. Our results show that the surface pool of NKCC1 T203/T207/T212A is decreased (by 1.1-fold) compared to that of the WT transporter ( Figure 6D,E). This decrease in the membrane stability of the transporter was accompanied by a 1.6-fold and a 1.7-fold decrease in the cluster density of NKCC1 T203/T207/T212A and NKCC1 T203/T207/T212/T217/T230A ( Figure 6F), compared to the WT. The remaining NKCC1 T203/T207/T212A and NKCC1 T203/T207/T212/T217/T230A clusters were not changed in size or fluorescence intensity compared to WT ( Figure 6G,H). This effect is reminiscent of that observed upon muscimol treatment (Figure 2) or WNK1 inhibition (Figure 4). Importantly, NKCC1 T203/T207/T212/T217/T230A prevented the muscimol-induced decrease in NKCC1 clustering ( Figure 6I-K). We conclude that GABA A R-mediated regulation of NKCC1 membrane, diffusion, clustering, and stability involves phosphorylation of its T203/T207/T212/T217/ T230 residues.

Functional Impact of NKCC1 Regulation by the WNK Pathway in Mature Hippocampal Neurons
Our work describes a regulation of membrane stability and clustering of NKCC1 by GABAergic activity. This regulation involves the phosphorylation of the transporter by the WNK/SPAK/OSR1 signaling pathway. To assess the functional relevance of this regulation, we looked at the impact of NKCC1 phosphorylation on [Cl − ] i . We used SuperClomeleon [25] to quantify potential changes in intracellular chloride concentration. Changes in concentration were inferred from changes in YFP/CFP ratios ( Figure 7A). This strategy has been shown to be effective in measuring variations in intracellular chloride levels in neurons. Indeed, we estimated in previous work the effect of gabazine and muscimol on [Cl − ] i showing a respective decrease and increase in chloride levels upon drug treatment [18]. As a control condition, we compared the YFP/CFP ratio of cells transfected with KCC2-WT vs. KCC2-T906/T1007E, i.e., a construction mimicking the phosphorylated state of the transporter with a reduced capacity to extrude chloride ions, notably by modifying its membrane stability and clustering [18,24]. An important decrease in the ratio was observed in cells transfected with KCC2-TE as compared to KCC2-WT, reflecting a higher [Cl − ] i ( Figure 7B). Thus, this approach allows the measurement of [Cl − ] i elevation resulting from manipulations of chloride cotransporter membrane expression and, thereby, function. We then tested whether the expression of endogenous or recombinant NKCC1 significantly impacted [Cl − ] i in our neuronal preparation. First, we determined the impact of an acute blockade of NKCC1 with bumetanide (5 µM) on the YFP/CFP ratio in neurons transfected with SuperClomeleon alone. The ratio was comparable in both bumetanide-exposed and non-bumetanide-exposed neurons ( Figure 7C). These results are in agreement with data suggesting that, at rest, endogenous NKCC1 does not significantly influence [Cl − ] i in mature hippocampal neurons [31,32]. Similarly, expression of the recombinant NKCC1-WT in mature neurons did not increase [Cl − ] i compared to neurons expressing the chloride probe alone, nor did it increase their sensitivity to bumetanide ( Figure 7C). This suggests that the neuron tightly regulates the level of recombinant NKCC1 present at the cell membrane.
If the expression of NKCC1-WT does not influence [Cl − ] i , then it is not surprising that a loss of function of the transporter cannot be detected. Indeed, we found that NKCC1-T203/T207/T212/T217/T230A, which shows a defect in membrane clustering compared to WT, has no effect on either the YFP/CFP ratio or the response to bumetanide ( Figure 7C). However, since NKCC1 plays an important role on [Cl − ] i in mature neurons under conditions where KCC2 is down-regulated [4,9,18,32], we performed additional analyses on neurons co-transfected with the mutant transporter KCC2-T906/1007E, which has a reduced chloride extrusion capacity [24] and Figure 7A. Overexpression of KCC2-T906/1007E was preferred to the expression of an shRNA against KCC2 because this strategy led to neuronal death when NKCC1 was expressed in concert. Interestingly, neuronal death was not observed when the shRNA against KCC2 was expressed (not shown). This indicates that the recombinant NKCC1 transporter is functional in neurons and that the influx of chloride through NKCC1 in neurons in the absence of chloride ion extrusion capacity is toxic. In agreement with previous works [31,32], this result also means that in mature hippocampal neurons, KCC2 is the major regulator of [Cl − ] i . However, no effect of the endogenous NKCC1 or of the recombinant NKCC1-WT or NKCC1-T203/T207/T212/T217/T230A was observed on the YFP/CFP ratio, nor on bumetanide sensitivity in conditions of low KCC2 activity ( Figure 7D). We concluded that at rest, in conditions of normal or reduced KCC2 expression, endogenous or exogenous NKCC1 transporters are not significantly contributing to [Cl − ] i in mature hippocampal cultured neurons.  7C). This suggests that the neuron tightly regulates the level of recombinant NKCC1 present at the cell membrane.  Since NKCC1 expression is increased in the adult epileptic brain and this expression contributes to increasing seizure susceptibility [3,7], we tested the contribution of the WNK1 signaling in the control of the membrane stability and function of NKCC1 in pathological conditions. For this purpose, neurons were acutely exposed to the convulsing agent 4-Amminopyridine (4-AP), a blocker of the voltage-dependent K + channels responsible for membrane repolarization. This experiment was performed in the absence of TTX + KYN + MCPG. The "4-AP condition" was compared to the control condition in the absence of any drug. We previously showed that acute exposure of hippocampal neurons to 4-AP induces KCC2 endocytosis [13]. Conversely, we show here that this treatment rapidly increases the membrane expression of NKCC1 ( Figure 7E). Pre-treatment of neurons with the inhibitor WNK463 prevented the 4-AP-induced increase in NKCC1 surface expression ( Figure 7E), implicating WNK1 in the upregulation of NKCC1 at the neuronal surface.
Although we have shown that NKCC1 does not participate in the regulation of intracellular chloride levels under basal activity conditions in mature hippocampal neurons ( Figure 7C,D), we show here the contribution of NKCC1 to neuronal chloride homeostasis under pathological conditions. Indeed, chloride imaging revealed that acute exposure to 4-AP induced a significant increase in intracellular chloride levels in neurons ( Figure 7F,G). This effect was blocked by pre-incubating the neurons with the WNK antagonist or by blocking NKCC1 activity with bumetanide ( Figure 7F,G), thus directly implicating the WNK signaling and NKCC1 in this regulation. Thus, we propose that 4-AP-induced hyperactivity activates the WNK pathway, which by phosphorylating NKCC1 on key threonine residues, increases its membrane expression and clustering, leading to intracellular chloride influx and decreased efficacy of GABAergic transmission.

Discussion
We studied, in mature hippocampal neurons, the cellular and molecular mechanisms regulating the cotransporter NKCC1, which transports chloride ions inside neurons. We showed that the transporter displays a heterogeneous distribution at the plasma membrane: it is either diffusely distributed and freely mobile in the membrane or organized in clusters where it is slowed down and confined. Here, we show that this distribution and behavior can be rapidly tuned by GABAergic activity changes. In particular, acute GABA A R activation or inhibition causes the escape of transporters from membrane clusters and their confinement in endocytic zones. GABA A R-mediated regulation of NKCC1 membrane distribution uses chloride as a secondary messenger and the Cl − -sensitive WNK1 pathway, which in turn affects the phosphorylation of key threonine residues on NKCC1. At rest, these modifications have little effect on [Cl − ] I , but they participate in Cl − accumulation in neurons in pathological conditions associated with an up-regulation of NKCC1 surface expression/function.
An increase in clustering generally correlates with a slowing down and confinement of the molecule in a sub-cellular compartment, e.g., synapses for neurotransmitter receptors. Conversely, a dispersion of molecules from clusters implies a lifting of the diffusion brakes. Usually, transmembrane proteins are confined in membrane clusters due to their binding to scaffolding proteins that anchor them to the cytoskeleton. However, they can escape from these confined regions by lateral diffusion. This is the case of excitatory glutamate receptors and inhibitory GABA A Rs [33,34], as well as of ion transporters such as the chloride cotransporter KCC2 [11,12]. However, we found that under basal activity conditions, a proportion of NKCC1 transporters are slowed down and confined to endocytic zones [35]. The confinement of NKCC1 in endocytic zones is further increased following neuronal exposure to gabazine or muscimol without leading to its internalization. This suggests that under certain activity conditions, NKCC1 would be stored in the endocytic zones from which it could, however, come out and diffuse into the membrane to join membrane clusters.
Acute blockade of glutamatergic activity by the TTX + KYN + MCPG drug cocktail confines NKCC1 to the axon [35]. This suggests that spontaneous glutamatergic activity conversely makes NKCC1 mobile along the axon. Here we investigated the role of GABAergic transmission on NKCC1 diffusion in the axon. We show that blocking GABA A R activity with gabazine in the presence of TTX + KYN + MCPG to prevent the indirect effects of gabazine on increased excitation removes NKCC1 diffusion constraints in the axon. Conversely, activation of GABA A R by muscimol in the presence of TTX + KYN + MCPG slows NKCC1 in the axon. Thus, GABAergic and glutamatergic transmission have opposite effects on NKCC1 diffusion in the axon. We propose that glutamatergic activity regulates by a homeostatic mechanism the axonal diffusion of NKCC1 to compensate for the increase in activity by decreasing the depolarizing/excitatory action of GABA A R in the axon. Here, muscimol activation of GABA A Rs confines NKCC1 in the axon to enhance the depolarizing action of GABA, which would potentially impact neurotransmitter release [36] and action potential firing [31].
The effects of gabazine on NKCC1 diffusion in the dendrite and axon are opposite, highlighting distinct regulatory mechanisms. In the case of NKCC1 regulation by glutamatergic transmission, different effects were also observed in the dendrite and axon [35]. Future experiments will determine whether this difference is due to variations in intracellular chloride concentration (with higher concentration in the axon than in the dendrite) and activation of the WNK pathway or to other mechanisms.
In the dendrites of mature neurons, we observed a similar effect of GABA A R activation or inhibition on the diffusion and clustering of dendritic NKCC1. In both cases, the transporter was sent to endocytic zones and confined there. The fact that there was no change in the global pool (surface + intracellular) of the transporter indicates that it is stored in the endocytic zones without being internalized and degraded. These endocytic zones have been shown to constitute reserve pools of neurotransmitter receptors, which can, depending on the synaptic demand, be released and reintegrated into the diffusing pool of receptors [37]. In the case of NKCC1, this reserve pool would allow a rapid increase in the transporter availability in the plasma membrane, for example, in pathological situations in which an up-regulation of NKCC1 membrane function has been observed [9].
By activating the GABA A R, muscimol raises [Cl − ] i [18]. An increase in [Cl − ] i blocks the activity of WNK1 and SPAK, OSR1 kinases [38,39], leading to the dephosphorylation of NKCC1-T203/207/212/217/230 [40][41][42] and reduced transporter activity [43,44]. We showed that following exposure to muscimol, NKCC1 escaped from membrane clusters and was confined in endocytic zones. Thus, the transition of the transporter between membrane clusters and endocytic zones by lateral diffusion would allow for modulating its availability in the membrane and its activity rapidly. The effects of muscimol are compatible with NKCC1-T203/207/212/217/230 dephosphorylation. Pharmacological (WNK-463 or closantel) or genetic (WNK-KD) blockade of the WNK pathway or the expression of NKCC1 TA3 or NKCC1 TA5 mutants that mimic NKCC1 dephosphorylation have the same effects as muscimol: they restrict NKCC1 in their movement and reduce the membrane clustering of the transporter. The demonstration that the effect of muscimol directly involves dephosphorylation of NKCC1-T203/207/212/217/230 was provided by the fact that the effect of muscimol on the clustering of NKCC1 can be prevented when the mutant NKCC1-TA5 was exposed to the drug, compared to WT.
In contrast, treatment of neurons with gabazine, by blocking the activity of GABA A Rs, decreases dendritic [Cl − ] i . In non-neuronal cells, low chloride activates WNK1/SPAK [38,39] by auto-phosphorylation of WNK1 S382 residue. Then, WNK phosphorylates in cascade SPAK on S373 and OSR1 on S325 [45] that in turn phosphorylate NKCC1 on T203/207/212/217/230 and increase the surface expression and activity of the transporter [44,46]. We showed that this signaling cascade is operant in mature hippocampal neurons: gabazine activates the WNK1/SPAK/OSR1 pathway by phosphorylation, thus inducing the phosphorylation of KCC2-T906/1007 as well as NKCC1-T203/207/212/217/230 [18]. If muscimol decreases NKCC1 clustering and confines it to endocytic zones, gabazine treatment should conversely induce the escape of the transporter from endocytic zones, thereby increasing its clustering and function in the membrane. Although we observed that a decrease in [Cl − ] i by substituting chloride with methane sulfonate decreases the confinement of NKCC1 and increases its membrane stability and clustering, the treatment of neurons with gabazine did not reproduce this effect. On the contrary, gabazine confined the transporter and induced the loss of its clustering just as muscimol did. However, we showed that antagonizing inhibition with gabazine reduces the surface expression of KCC2 by increasing the lateral diffusion and internalization of the transporter [18]. This results in reduced intracellular chloride extrusion capacity of the neuron leading to a significant increase in [Cl − ] i as monitored by SuperClomeleon imaging [18]. KCC2 is more effective in regulating [Cl-] i than NKCC1 in dendrites of mature neurons [47,48]. We, therefore, hypothesize that the regulation of NKCC1 diffusion-capture by gabazine is not due to a reduction in [Cl − ] i but instead to an increase in [Cl − ] i following KCC2 regulation by the WNK1/SPAK/OSR1 pathway. Thus, changes in the membrane expression of KCC2 (under the control of the WNK1 pathway) would condition that of NKCC1, thus allowing [Cl − ] i to be maintained at a low level in mature neurons. This would explain why expressing recombinant NKCC1-WT in mature neurons does not significantly increase its expression at the membrane, nor does it increase [Cl − ] i . This suggests that the membrane expression level of NKCC1 is under the control of KCC2 and its tuning of [Cl − ] i .
Nevertheless, we showed that the membrane stability and clustering of NKCC1 can be rapidly regulated by lateral diffusion and that this mechanism is rapidly controlled by GABAergic inhibition and the WN1K/SPAK/OSR1 pathway on the dendrites of mature neurons. Although this pathway has little influence on the amount/function of NKCC1 at the neuronal surface under basal activity conditions, we propose that it may play a role in pathological situations associated with increased expression levels of NKCC1 based on our 4-AP data. Interestingly, in the pathology, the upregulation of NKCC1 is often accompanied by a down-regulation of KCC2 at the neuronal surface [3,16]. KCC2 is also regulated by diffusion-capture. We showed that a short exposure of neurons to the convulsive agent 4-AP increases the lateral diffusion of KCC2, which escapes from the clusters, is internalized and degraded [13]. Thus, lateral diffusion would be a general mechanism to control the membrane stability of chloride cotransporters. Moreover, the fact that KCC2 is also regulated in mature neurons by the WNK1 pathway [18] and that this regulation has an inverse effect on membrane stability, clustering, and function of KCC2 indicates that this pathway is a target of interest in the pathology. Inhibition of the pathway would prevent the loss of KCC2 and the increase in NKCC1 at the surface of the neuron, thus preventing the abnormal rise of [Cl − ] i in the pathology and the resulting adverse effects.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/cells12030464/s1, Figure S1: NKCC1 diffusion in the axon is increased upon GABA A R blockade.

Data Availability Statement:
The data that support the findings of this study are available from the corresponding author upon reasonable request. The transfer of plasmids generated for this study will be made available upon request. A Materials Transfer Agreement may be required.