NKCC-1 mediated Cl− uptake in immature CA3 pyramidal neurons is sufficient to compensate phasic GABAergic inputs

Activation of GABAA receptors causes in immature neurons a functionally relevant decrease in the intracellular Cl− concentration ([Cl−]i), a process termed ionic plasticity. Amount and duration of ionic plasticity depends on kinetic properties of [Cl−]i homeostasis. In order to characterize the capacity of Cl− accumulation and to quantify the effect of persistent GABAergic activity on [Cl−]i, we performed gramicidin-perforated patch-clamp recordings from CA3 pyramidal neurons of immature (postnatal day 4–7) rat hippocampal slices. These experiments revealed that inhibition of NKCC1 decreased [Cl−]i toward passive distribution with a time constant of 381 s. In contrast, active Cl− accumulation occurred with a time constant of 155 s, corresponding to a rate of 15.4 µM/s. Inhibition of phasic GABAergic activity had no significant effect on steady state [Cl−]i. Inhibition of tonic GABAergic currents induced a significant [Cl−]i increase by 1.6 mM, while activation of tonic extrasynaptic GABAA receptors with THIP significantly reduced [Cl−]i.. Simulations of neuronal [Cl−]i homeostasis supported the observation, that basal levels of synaptic GABAergic activation do not affect [Cl−]i. In summary, these results indicate that active Cl−-uptake in immature hippocampal neurons is sufficient to maintain stable [Cl−]i at basal levels of phasic and to some extent also to compensate tonic GABAergic activity.


Steady-state distribution of [Cl − ] i in immature CA3 pyramidal cells.
In this study we recorded in total from 121 CA3 pyramidal cells under gramicidin-perforated patch-clamp conditions. We estimated the reversal potential of GABAergic (E GABA ) and glycinergic (E Gly ) currents from short (2-10 ms) puffs of 30 µM muscimol or 0.2-1 mM glycine applied focally to the soma of the pyramidal cells (Fig. 1A). The use of glycine pulses was necessary for the determination of [Cl − ] i in part of the following experiments, as in these experiments gabazine or picrotoxin were used to eliminate phasic and tonic GABAergic currents. At a holding potential of − 70 mV these cells showed an E GABA  In order to determine the kinetics of passive Cl − -efflux and to confirm that this high [Cl − ] i was maintained by the activity of the NKCC1 14,35 , we first analyzed the effect of the NKCC1 inhibitor bumetanide on [Cl − ] i . Bath application of 10 µM bumetanide resulted in a significant (p = 0.028, Wilcoxon) decline of [Cl − ] i from 14.5 [11.5, 21.6] mM (n = 6) to 10.2 [9.0, 11.8] mM (n = 6) within ~ 10 min (Fig. 1D,E). This decline in the [Cl − ] i could be fitted by a monoexponential function using a τ of 381 s (Fig. 1D). From this function we estimated a maximal passive Cl − efflux of 15.5 µM/s at a [Cl − ] i of ~ 15 mM. The [Cl − ] i of 10.2 [9.0, 11.8] mM (n = 6) obtained in the presence of bumetanide is in the range of the passive Cl − distribution of 9.1 mM. In summary, this result confirms that NKCC1 substantially contribute to the active Cl − accumulation and is counteracting a passive Cl − efflux 5,8 . Estimation of the capacity of NKCC1 mediated Cl − uptake. Since the kinetics of [Cl − ] transport is one major factor influencing ionic plasticity 17,[26][27][28]39,40 , we next determined the kinetic properties of Cl − uptake after an artificial [Cl − ] i reduction. To quantify the kinetics of the [Cl − ] i reuptake we decreased [Cl − ] i by 25 pulses of either 1-3 mM glycine or 30 µM muscimol applied with a frequency of 0.5 Hz to voltage-clamped neurons ( Fig. 2A). This procedure significantly (p = 0.002, Wilcoxon) reduced the [Cl − ] i by 3.2 [2.6, 3.8] mM (n = 12) from 11.4 [10.3, 11.8] mM to 7.9 [7.7, 8.4] mM (Fig. 2B). As neither the amount of the [Cl − ] i decrease (p = 0.808, Mann-Whitney) nor the time constants of [Cl − ] i recovery (p = 0.291, Mann-Whitney) were significantly different between glycine-and muscimol-application experiments, the data was pooled. The subsequent recovery of [Cl − ] i was monitored by determining E REV with a small number of test pulses given at intervals of ~ 100 s, to avoid substantial Cl − fluxes by these test pulses 4 . These experiments showed that [Cl − ] i returned to the resting values within ~ 10 min (Fig. 2C). This increase in [Cl − ] i could be described with a monoexponential function using a time constant τ of 155 s. At a [Cl − ] i of 9.1 mM, which represents a passive distribution at − 70 mV holding potential and thus eliminates passive fluxes, the active Cl − uptake rate amounted to 15.4 µM/s. In summary, these results indicate that NKCC1-mediated Cl − uptake is sufficient to maintain [Cl − ] i at the observed values, but that this transport process is rather slow in immature CA3 pyramidal neurons.
Effect of spontaneous phasic GABAergic activity. Given (Fig. 3B). Note that the tendency (p = 0.07, Mann-Whitney U-test) to higher basal [Cl − ] i in these neurons, as compared to all recordings, led to a higher driving force for Cl − ions and would thus result in even higher activity-dependent [Cl − ] i changes. We conclude from these results that the capacity of NKCC1-mediated Cl − uptake in immature CA3 pyramidal neurons is sufficient to cope with the Cl − -influx caused by spontaneous GABAergic synaptic inputs.
Tonic currents mediated by extrasynaptic GABA receptors contribute substantially to passive [Cl − ] i fluxes in immature neurons 4 , as such tonic currents mediate a larger charge transfer 43 . Since in the immature hippocampus extrasynaptic receptors substantially contribute to the excitability 9,13,41 , we also investigated whether tonic GABAergic currents influence [Cl − ] i . For this purpose, we blocked tonic and phasic GABAergic currents  To simulate phasic GABAergic activity, we modeled stochastic activation of 107 GABA A synapses located randomly in the soma and perisomatic dendrites 45 (Fig. 4D). The initial values for the conductance of GABA A synapses (g GABA = 169 pS) and the frequency of GABAergic synaptic currents (2.14 Hz) were based on the mean values of the experimental data. In accordance with the patch-clamp observation, addition of physiological levels of GABAergic synaptic inputs had only a marginal effect on [Cl − ] i . After 100 s of continuous GABAergic activity at 2.14 Hz [Cl − ] i decreased by only 0.012 mM (Fig. 4E,F). Augmenting g GABA enhanced the synaptically evoked [Cl − ] i decline (Fig. 4F), which however remained small (0.114 mM) even if g GABA was increased to 1.69 nS. Increasing the frequency of GABAergic synaptic inputs from 2.14 to 5.35 Hz had only a marginal effect of To analyze the influence of tonic GABAergic currents, we omitted the tonic GABAergic conductance of 8.75 nS/cm 2 , which changed the holding current (I hold ) by 0.8 pA and led to a [Cl − ] i increase by 0.32 mM (Fig. 4H). Enhancing tonic GABAergic currents by adding multiples of 5 nS/cm 2 to the basal tonic conductance induced a dose-dependent decrease in [Cl − ] i that, however, remained below 0.8 mM (Fig. 4I). These additional tonic conductances between 5 and 25 nS/cm 2 induced a linear decrease in the inward current between − 0.3 and − 1.28 pA, respectively.
In summary, these simulations with a realistic computational model of Cl − dynamics support the observation, that NKCC1-mediated Cl − transport is sufficient to maintain [Cl − ] i at basal levels of phasic GABAergic activity, while tonic currents have a mild effect on steady-state [Cl − ] i .   38 . While published permeability ratios for GABA A (0.44) and glycine receptors (0.40) in hippocampal neurons are available 46 , using these values resulted in unrealistically low [Cl − ] i values (ranging from 0.4 to 30 mM). Therefore we prefer to use the values provided by Bormann et al. 38 . Importantly, the qualitative results of our study will not be altered, if the higher relative HCO 3 − permeability ratios are used. The first outcome of this study is that inhibition of NKCC1 with 10 µM bumetanide induced a decline in [Cl − ] i towards the passive distribution, which supports the fact that this transporter constitutes the main Cl − uptake mechanism in immature CA3 pyramidal cells 14,35,47 . Bumetanide has been reported to also inhibit other targets, like KCC2 or aquaporins 48,49 , however, at slightly higher effective doses of 55 µM and 100 µM, respectively. An inhibition of these transporters will have no effect  19 , and to neurons from juvenile hippocampal slice cultures, where ongoing phasic activity is required to allow dynamic changes in E GABA 51 . However, in the present simulations [Cl − ] i was determined in the soma, to match the modelling results with the experimental design of the gramicidin perforated patch experiments. Within the dendritic compartment these levels of GABAergic synaptic activity will induce substantial GABAergic [Cl − ] i transients 17 .
In addition, the present study demonstrated that inhibition of both, phasic and tonic GABAergic currents with picrotoxin 13 evoked a small, but significant [Cl − ] i increase. In line with this observation, the pharmacological induction of extrasynaptic GABAergic currents with THIP led to a substantial reduction in [Cl − ] i . This observation could be replicated in the NEURON simulations by the addition of tonic Cl − conductances. Our experiments suggest that the substantial tonic current found in immature hippocampal neurons 52 , contributes to steady state [Cl − ] i . The dissimilar effects of phasic and tonic GABAergic currents directly reflect the different levels of charge transfer mediated by the short phasic responses (ca. 0.015 pC/s at 2.13 Hz and a g GABA of 169 pS) in contrast to the persistent but small amplitude tonic GABAergic currents (ca. 0.46 pC/s at 8.75 nS/cm 2 ) 43 .
A variety of studies in the adult brain have shown that excessive GABAergic stimulation can induce Cl − -fluxes that exceed the capacity of active Cl − transport and consequently increase [Cl − ] i and alter GABAergic responses 15,16,18,25,40,51,53 . The low capacity of the NKCC1-mediated Cl − uptake makes immature neurons particularly susceptible to such effects. On the other hand, spontaneous, highly correlated activity transients are typical Scientific Reports | (2020) 10:18399 | https://doi.org/10.1038/s41598-020-75382-1 www.nature.com/scientificreports/ for developing neuronal systems 54 . These correlated activity transients are characterized by the synchronous activation of glutamatergic and GABAergic activity 55,56 . The resulting massive GABAergic inputs may exceed the capacity of transmembrane Cl − transport and induce substantial [Cl − ] i changes. The depolarizing glutamatergic inputs during correlated network activity can even augment the GABAergic Cl − fluxes and thus aggravate the [Cl − ] i alterations 57 . Indeed it has been demonstrated that such physiological bursts of activity can led to substantial [Cl − ] i shifts in the spinal cord 33,58 , neocortex 19 , and immature hippocampus 17 . Whereas in the adult system such changes will decrease GABAergic inhibition and thus contribute to the establishment of hyperexcitable states and reduced pharmacological responsiveness 18 , in immature neurons the activity-dependent [Cl − ] i reduction will decrease the excitatory potential of GABA A receptor-mediated responses. And since shunting inhibition remains constant 55,59 , this activity-dependent [Cl − ] i decrease will augment the inhibitory potential of GABAergic responses. Therefore, the low capacity of Cl − export in immature neurons may be an adaptation to prevent hyper-excitability mediated by depolarizing GABA responses, as has been originally suggested by Ben-Ari 3 .
In addition, at least for the immature spinal cord it has been demonstrated that such activity-dependent [Cl − ] i transients can determine the frequency of spontaneous activity transients by temporarily reducing the excitatory effect of GABA 58 . Thus it is tempting to speculate that recurrent alterations in [Cl − ] i may also contribute to slow oscillatory phenomena in other regions of the developing nervous system. In summary, the results of our present study demonstrate that the capacity of Cl − accumulation is limited in immature hippocampal CA3 pyramidal neurons. This finding, in combination with a quantification in immature cortical neurons 4 and the observation of activity-related [Cl − ] i transients in other immature tissues 19,33 , suggests that an unstable [Cl − ] i homeostasis may be an innate feature of immature neurons.

Methods
Slice preparation. All experiments were conducted in accordance with EU directive 86/609/EEC for the use of animals in research and the NIH Guide for the Care and Use of Laboratory Animals, and were approved by the local ethical committee (Landesuntersuchungsanstalt RLP, Koblenz, Germany). All efforts were made to minimize the number of animals and their suffering. Wistar rat pups of P4-7 were obtained from the local breeding facility and were deeply anesthetized with enflurane (Ethrane, Abbot Laboratories, Wiesbaden, Germany). After decapitation, the brains were quickly removed and immersed for 2-3 min in ice-cold standard artificial cerebrospinal fluid (ACSF, composition see below). Coronal slices (400 µm thickness) including the hippocampus were cut on a vibratome (HR2, Sigmann Elektronik, Hüffenhardt, Germany). The slices were stored in an incubation chamber filled with oxygenated ACSF at room temperature before they were transferred to the recording chamber.
Data acquisition and analysis. Gramicidin-perforated whole-cell patch-clamp recordings were performed as described previously 4,13 at 31 ± 1 °C in a submerged-type recording chamber attached to the fixed stage of a microscope (BX51 WI, Olympus). Pyramidal neurons in stratum pyramidale of the CA3 region were identified by their location and morphological appearance in infrared differential interference contrast image. Patchpipettes (5-12 MΩ) were pulled from borosilicate glass capillaries (2.0 mm outside, 1.16 mm inside diameter, Science Products, Hofheim, Germany) on a vertical puller (PP-830, Narishige) and filled with pipette solution containing 130 KCl, 1 CaCl 2 , 2 MgCl 2 , 11 EGTA, 10 HEPES, 2 Na2-ATP, 0.5 Na-GTP (pH adjusted to 7.4 with KOH and osmolarity to 306 mOsm with sucrose). For gramicidin-perforated patch-clamp recordings 10-50 µg/ ml gramicidin D (Sigma, St Louis, MO, USA) was added from a stock solution (1-2 mg/ml in DMSO) on the day of experiment. Experiments were omitted from analysis, when an instantaneous or constant shift in the muscimol/glycine reversal potential towards positive values, determined by the high [Cl − ] of the pipette solution, occurred, as they indicate insufficient perforated-patch conditions. Signals were recorded with a discontinuous voltage-clamp/current-clamp amplifier (SEC05L, NPI, Tamm, Germany), low-pass filtered at 3 kHz and stored and analyzed using an ITC-1600 AD/DA board (HEKA) and TIDA software. Input resistance and capacitance were determined from a series of hyperpolarizing current steps. The apparent cell surface was estimated using a specific capacitance of 2 µF/cm 2 . Spontaneous postsynaptic currents (sPSCs) were detected and analysed from whole-cell patch-clamp recordings according to their amplitude and shape by appropriate settings using Minianalysis Software (Synaptosoft, Fort Lee, NJ).
The  38 . GABAergic and glycinergic currents were evoked by brief (2-10 ms) pulses of 30 µM muscimol or 0.2-1 mM glycine from a patch pipette positioned close to the soma via a custom built pressure application system (Lee, Westbrook, CT) at a pressure of 0.5 bar. The use of glycine pulses was necessary to allow the determination of [Cl − ] i in the presence of gabazine or picrotoxin, which eliminate GABAergic currents 13 .
All values were given as median ± interquartile range, in the panels median ± interquartile range was used for time-dependent plots, while summarized results were shown as box and whisker plots (minimum, first quartile, median, third quartile, maximum). For statistical analysis of unpaired data Mann-Whitney U-tests and for paired data Wilcoxon signed-rank test were used (Systat 11, Point Richmond, CA). Significance was assigned at levels of 0.05 (*), 0.01 (**) and 0.001 (***).  60 and expanded the x-/y-dimensions by 12.5% and the z-dimension by 50%.
Compartmental modeling. The reconstructed CA3 pyramidal cell (see above) was imported into the NEURON simulation program (neuron.yale.edu). The following passive parameters were used: R a (specific axial resistance) = 35.4 Ωcm; g pas (passive specific membrane conductance) = 17.05 nS/cm 2 ; E pas = -74.05 mV, C m (specific membrane capacitance) = 1 µF/cm 2 . In addition, a tonic leak Cl − conductance: with a conductance g tonic of 8.75 nS/cm 2 was inserted. Implementing these parameters in the reconstructed morphology resulted in a resting membrane potential of − 70 mV and an input resistance of 306 MΩ. Cl − diffusion and uptake were calculated by standard compartmental diffusion modeling 40,44 . To simulate intracellular Cl − dynamics, we adapted our previously published model 40 . Longitudinal Cl − diffusion along dendrites was modeled as the exchange of Cl − between adjacent compartments. For radial diffusion, the volume was discretized into a series of 4 concentric shells around a cylindrical core and Cl − was allowed to flow between adjacent shells 61 . The free diffusion coefficient of Cl − inside neurons (D Cl ) was set to 2 µm 2 /ms 53 . To simulate Cl − uptake, a pump mechanism for transmembrane Cl − transport was included. Cl − transport was modeled as exponential recovery of [Cl − ] i to its target [Cl − ] i ([Cl − ] i 0 ) with a time constant τ Cl .
The pump mechanism approximates an NKCC1-like Cl − transport mechanism. The impact of GABAergic Cl − currents on [Cl − ] i was calculated as: with F = 96,485 C/mol (Faraday constant). GABA A synapses were simulated as a postsynaptic parallel Cl − and HCO 3 − conductance with exponential rise and exponential decay 40 : where P is a fractional ionic conductance that was used to split the GABA A conductance (g GABA ) into Cl − and HCO 3 − conductance. E Cl and E HCO3 were calculated from Nernst equation. The GABA A conductance was modeled using a two-term exponential function, using values of rise time (0.5 ms) and decay time (37 ms) 17 38 . The GABA A inputs (107 synapses, peak conductance 0.169 nS) 17 were activated stochastically (Poisson) with a frequency of 0.02 Hz, corresponding to a main PSC frequency of 2.14, except where noted. Source codes of all models are available at ModelDB (https ://model db.yale.edu/26681 1; password is "hippocampus").