The study was performed on E18.5 embryos of Kcc2^+/+ and Kcc2^−/− mice as characterized previously (Hubner et al., 2001). The animal care and handling was performed in accordance with the guidelines of the European Union Council and the INSERM regulations on the use of laboratory animals.

Intact hippocampi were prepared as described previously (Khalilov et al., 1997). After cervical dislocation, the brain was removed quickly and submerged in ice-cold oxygenated (95% O₂/5% CO₂), choline-containing artificial CSF (ACSF) cutting solution, (in mM): 110 choline chloride, 2.5 KCl, 1.25 NaH₂PO₄, 0.5 CaCl₂, 7 MgCl₂, 25 NaHCO₃, 7 d-glucose, pH 7.4. The hemispheres were separated, and after removing the cerebellum, the frontal part of the neocortex, and surrounding structures, intact hippocampi were dissected from the septohippocampal complex and transferred into a beaker containing oxygenated ACSF containing (in mM):126 NaCl, 3.5 KCl, 2.0 CaCl₂, 1.3 MgCl₂, 25 NaHCO₃, 1.2 NaH₂PO₄, and 11 d-glucose (pH 7.4) and incubated at least 1 h before use. The intact hippocampi were placed into the conventional fully submerged chamber and superfused with oxygenated ACSF (30–32°C, 5–6 ml/min). The hippocampi were fixed by entomological needles to the sylgard-covered bottom.

Extracellular field potentials and multi unit activities (MUA) were recorded in the intact in vitro hippocampal preparations using tungsten wire electrodes (diameter: 50 μm, California Fine Wire, Grover Beach, CA, USA) and a low-noise multichannel DAM-8A amplifiers (WPI, GB; low filter: 0.1 Hz; high filter: 3 kHz; ×1000). Electrical stimulations were performed with a bipolar electrode (10–20 V, 40 μs). The signals were digitized using an analog-to-digital converter (Digidata 1440A, Axon Instruments, USA). pCLAMP 10.0.1.10 and Clampfit 10.1 (Axon Instruments, USA), MiniAnalysis 6.03 (Synaptosoft, Decatur, CA, USA) and Origin 7.5 (Microcal Software, USA) programs were used for the acquisition and analysis of Extracellular field potentials and MUA.

Hippocampal transverse slices (300 μm) from the middle portion of each hippocampus were cut with a vibrating microtome (Leica VT 1000S, Germany) in ice-cold oxygenated (95% O₂/5% CO₂), choline-replaced ACSF and were incubated at room temperature in ACSF. Slices were allowed to recover for at least 90 min before recording.

Whole-cell recordings were performed from CA3 area in voltage clamp mode at different holding potential using Axopatch 200B (Axon Instruments, USA). The whole-cell patch pipettes had a resistance of 6–8 MΩ when filled with solution containing (in mM): 110 cesium methanesulfonate, 20 CsCl, 10 HEPES, 2 MgCl₂, 1 EGTA, 10 Na-phosphocreatine, 4 ATP–Mg, and 0.4 GTP–Na. The pH of the intracellular solutions was adjusted to 7.2 and the osmolarity to 280–290 mOsmol l⁻¹. The access resistance (15–30 MΩ) and input resistance (1–2 GΩ) were continuously monitored throughout the experiment. Occasional recordings with higher access resistance, lower input resistance, or those that displayed more than 20% changes in access and input resistances were discarded from analysis. External stimuli were applied using bipolar electrodes placed in the stratum radiatum (CA3 region). Stimulation intensity was adjusted during the first min of whole-cell recording to evoke 10–20 pA responses.

Gramicidin-perforated patch-clamp recordings were performed as described previously (Tyzio et al., 2003).

Drugs used were purchased from Sigma, (4-aminopyridine, 4-AP) or Tocris (bicuculline, CNQX, D-APV).

The expression of KCC2 was studied by immunohistochemistry. Brains from E18.5 mice were fixed in paraformaldehyde 4% for 24 h, then embedded in optical cutting temperature (OCT) compound (Tissue-Tek, Sakura Finetek Inc., NL, USA) before serially cut, in 15 μm sections, with a cryostat. Selected sections were processed for immunocytochemistry. They were first incubated for 2 h, at room temperature, with a mixture of 5% normal goat serum (NGS) in PBS with 0.3% Triton X100. They were then, incubated overnight at 4°C with the primary antibody diluted in PBS with 1% NGS and 0.1% Triton X100. After rinsing in PBS, sections were incubated with the fluorescent-labeled secondary antibody (1/1000, Chemicon, France) for 1 h at room temperature. The primary antibodies were an anti-KCC2 antibody (rabbit; dilution 1/1000; US Biological, Euromedex, France) or an anti-synaptophysin antibody (mouse, dilution 1/150, Chemicon, France).

Images were taken using a Zeiss confocal microscope (Wetzlar, Germany). The quantitative analysis of the fluorescence was performed in blind using Metamorphe software (Roper Scientific, USA). For analysis, on each image we manually drew regions comprising stained tissue and excluding nuclei and casual irregularities (empty spaces). The threshold of the detection of fluorescence clusters was set as Mean + 3 SD of the fluorescence level in Kcc2^−/− sections. The classification of the KCC2-positive neurons to groups “cytoplasm” or “membrane region” was done manually relying on the detectable fluorescence band on the periphery or center of the cell. All acquisitions and analysis were done blind.

Field recordings of ongoing neuronal activity in intact hippocampi isolated from E18.5 Kcc2^+/+ and Kcc2^−/− embryos revealed major differences between genotypes. In accordance with earlier studies, wild-type hippocampi generated the characteristic network recurrent population events that were reminiscent of giant depolarizing potentials (GDPs; Ben-Ari et al., 1989, 2007; Khalilov et al., 1999). In striking contrast, in Kcc2^−/− hippocampi, the spontaneous network events were significantly more frequent, longer-lasting, had higher amplitudes, and contained more spikes (Figures 1A,B). Another difference was that while in wild-type hippocampi 98.3% of spontaneous events had characteristic bi-phasic form (examples in Figures 1A and 2A, six experiments, 10 analyzed events per experiment), in Kcc2^−/− hippocampi the majority of spontaneous events (72.7%, n = 5, 30 events per experiment) had sharp tri-phasic form (see example in Figure 2B). These electrical activities were similar to epileptiform events observed in epileptic tissue (Cohen et al., 2002; Khalilov et al., 2003; Nardou et al., 2009, 2011; Wittner et al., 2009). Consistent with different spontaneous activities, evoked field potentials in Kcc2^−/− hippocampi differed clearly from those evoked in Kcc2^+/+ structures (Figures 1A,B).

To characterize the contribution of GABA-ergic and glutamatergic neurotransmission into generation of the spontaneous network activities we first exposed acute hippocampi to a CNQX, a blocker of AMPA receptors, and found that it fully abolished spontaneous events in both Kcc2^+/+ and Kcc2^−/− hippocampi (not shown, n = 5 per condition). By contrast, the application of Bicuculline, a GABA_A receptor antagonist, exerted different effects on Kcc2^+/+ and Kcc2^−/− hippocampi. In wild-type structure it produced a transient (4–5 min) inhibition of the spontaneous GDPs (Figure 2A). During first 4 min of the exposure to bicuculline the frequency decreased more than fourfold from (3.5 ± 0.5) × 10⁻³ Hz to (0.6 ± 0.1) × 10⁻³ Hz, n = 6. This is in keeping with the important contribution of depolarizing GABA in their generation (see Ben-Ari et al., 1989; Khalilov et al., 1999; Ben-Ari et al., 2007; Cherubini et al., 2011). However, within longer time, the bicuculline led to the generation of interictal-like events (IILE) [(6.8 ± 0.7) × 10⁻³ Hz, n = 6] reflecting the dual action of GABA, a depolarization, and a shunting action that, when blocked, enhances neuronal excitability (Khalilov et al., 1999). In contrast to naïve hippocampi, the ongoing spontaneous network activities recorded in Kcc2^−/− hippocampi were fully blocked by bicuculline even during long-lasting applications. As shown in Figure 2B, ongoing activity did not re-appear even after >20 min of application (n = 5). To unravel synchronized activities, we applied the K⁺ channel blocker 4-aminopyridine (4-AP, 100 μM, n = 6) that triggered IILEs similar to those observed spontaneously prior to the application of bicuculline (Figure 2B). Thus, embryonic Kcc2^−/− hippocampi generate spontaneous events that require depolarizing/excitatory actions of GABA as they are fully blocked by GABA receptor antagonists. Yet, the properties of these events are reminiscent of events recorded in epileptic tissues (Cohen et al., 2002; Khalilov et al., 2003). Understanding of the cellular and molecular perturbations leading to change of the network properties in embryonic Kcc2^−/− hippocampi and in epileptic tissue is an important subject for separate study.

To determine the alterations of GABA and glutamatergic postsynaptic currents (PSCs) in Kcc2^−/− in comparison to naïve neurons, we used whole-cell recordings from CA3 pyramidal neurons. We showed previously that in immature rat hippocampus most of CA1 pyramidal neurons are silent (Tyzio et al., 1999). In agreement with this, 14 out of 19 neurons recorded in E18.5 Kcc2^+/+ mice hippocampi were fully silent, generating neither spontaneous nor evoked PSCs (Figure 3A₁), four neurons showed GABA PSCs (Figure 3A_3,4), one neuron generated both GABA and glutamate PSCs but none had only glutamate PSCs as summarized in Figure 3D. In striking contrast, in Kcc2^−/− slices most neurons were already active at the same age; only 5 out of 22 recorded neurons were silent, 17 neurons showed spontaneous GABA-ergic and/or glutamatergic PSCs (Figures 3B,D). Also, electrical stimuli generated long-lasting multisynaptic responses in Kcc2^−/− slices that were never observed in Kcc2^+/+ slices (n = 19; Figure 3C).

Consistent with the electrophysiology data, immunolabeling with anti-synaptophysin antibody revealed significant differences between Kcc2^−/− and Kcc2^+/+ hippocampal sections (Figure 4). Blind analysis revealed that Kcc2 silencing induced a sixfold increase in the density of synaptophysin-positive clusters in CA3 pyramidal layer and twofold increase in CA3 stratum radiatum layer (Figure 4B, left panel). Moreover, in addition to increase of cluster density, there was also a significant increase of cluster brightness in Kcc2^−/− sections (2.8-fold in pyramidal layer and 1.6-fold in radiatum layer; Figure 4B, right panel) indicating an enhanced synapse formation in Kcc2^−/− versus Kcc2^+/+ mice. Thus, KCC2 controls the development of synaptic connections in the embryonic mice hippocampi.

As shown in Figure 5A, KCC2 was clearly detectable in CA3 pyramidal region of Kcc2^+/+ already 1 day before birth (E18.5), but not in Kcc2^−/− hippocampi. In E18.5 sections the average intensity of KCC2 immunofluorescence in both CA3 pyramidal and radiatum layers was at least twofold higher than the background fluorescence measured in Kcc2^−/− sections and constituted approximately 50% of the KCC2 immunofluorescence measured in sections obtained from older (P7) Kcc2^+/+ animals (Figure 5B). At E18.5 the KCC2-positive immunofluorescence was detected in 13.1 ± 1.9% (n = 3) of CA3 pyramidal neurons (Figure 5C). In the remaining 86.9 ± 1.9%, the level of KCC2 immunofluorescence was within the range of background variability (<mean + 3 SD; Figure 5C). Among KCC2-positive neurons, in the majority of cells the labeling was localized in the cytoplasm (Figure 5A, insert i1). Only 17.8% of KCC2-positive neurons (corresponding to 2.3 ± 0.5% of all analyzed neurons) included KCC2 in membrane region (Figure 5A, insert i2). Consistent with previous reports (Rivera et al., 1999; Stein et al., 2004), the expression of the KCC2 increased during development: at P7 82.7 ± 0.9% neurons were KCC2-positive and 79.3% of these neurons expressed KCC2 in their membrane region (Figure 5A, insert i4 and Figure 5C).

Thus, already during embryonic hippocampal development, there is an expression of the KCC2 in individual neurons of CA3 pyramidal layer that is mainly localized in the cytoplasm. Some neurons at this stage of development start to express KCC2 in membrane region.

Earlier studies using perforated patch-clamp recordings [at P1–P3 (Stein et al., 2004)] or single channel analysis (Tyzio et al., 2007) have shown that E_GABA was strongly depolarizing in embryonic CA3 and CA1 pyramidal neurons, respectively. Using gramicidine-perforated patch-clamp recordings, we did not find a significant difference between E_GABA recorded from CA3 neurons of Kcc2^+/+ and Kcc2^−/− embryonic hippocampi (P = 0.87, Figure 5D) confirming that [Cl⁻]_i is not regulated at this early developmental stage by KCC2. It is however possible that in naïve neurons KCC2 plays an important role in a small subpopulation of neurons (see Discussion).

There is a general consensus that the expression of the KCC2 starts during postnatal stages of the development of hippocampi in rat and mice (Stein et al., 2004; Blaesse et al., 2006; Zhu et al., 2008). This also implies a progressive externalization of the co-transporter possibly mediated by a developmental sequence of the phosphorylation mechanisms recently reported (Rinehart et al., 2009; Lee et al., 2010). Our observations of a similar E_GABA in Kcc2^−/− and Kcc2^+/+ is in keeping with this suggestion and with the largely depolarizing action of GABA at early stages (Ben-Ari et al., 2007). A recent study elegantly showed that a regulatory action of KCC2 in immature neurons might indeed develop through an ion transport-independent mechanism, as the overexpression of inactive N-terminus KCC2 deleted construct produced similar to wild-type KCC2 effects (Horn et al., 2010). Similar results, but different conclusions concerning the mechanisms of KCC2 action, were obtained by Cancedda et al. (2007) who showed that viral overexpression of the KCC2 in immature cortical neurons slows-down neuronal maturation. Our work is the first study showing that endogenous KCC2 and not only artificially expressed KCC2, regulate maturation in immature neurons.

Can we completely exclude the contribution of chloride shifts in the effects of KCC2 ablation? The finding that a GABA receptor antagonist generates epileptiform activities in naïve neurons – after long-lasting applications – but block fully ongoing activities generated by Kcc2^−/− embryos is intriguing as it suggests that excitatory actions of GABA are enhanced in the Kcc2 deficient in comparison to control mice. This stands in apparent contradiction with the similar E_GABA in Kcc2^−/− and Kcc2^+/+. However, there are many observations showing that chloride levels play important roles at embryonic stages in selective cell populations. Thus, in developing neocortex, while the majority of pyramidal cells and interneurons are KCC2-negative, some interneurons express high amount of KCC2 and the activity of this transporter orchestrates interneuron migration (Bortone and Polleux, 2009). In keeping with this, bicuculline alters neuronal migration in vitro (Manent et al., 2005). In the present work we report an existence of small population of neurons with KCC2 expressed in membrane region (Figure 5). Thus, one chloride-based alternative possibility is that the loss of the co-transporter affects a subpopulation of neurons impacting the entire network. One more possibility could be that naïve KCC2-positive neurons exhibit difference in [Cl⁻]_i homeostasis on the level of dendrites, but not soma. Studies show the existence of KCC2-dependent somato-dendritic gradient of chloride in mature hippocampal neurons (Khirug et al., 2008; Pellegrino et al., 2011). Testing these hypotheses would require the development of appropriate models allowing [Cl⁻]_i measurement in different neuronal compartments as well as suppression of the KCC2 in different neuron subpopulations during embryonic development.

Our results show clearly that ablation of KCC2 enhances synaptogenesis in the immature hippocampus, i.e., endogenous KCC2 inhibits formation of the synaptic connections and neuronal network activity. This finding is in a good agreement with previous reports by Cancedda et al. (2007) and Horn et al. (2010) illustrating that in vivo overexpression of KCC2 in immature neurons slows-down neuronal development. Interestingly, all these results are in apparent contradiction with previous reports performed on primary neuronal cultures and showing that KCC2 is required for formation of GABA and glutamate synaptic connections. Particularly, we found that overexpression of KCC2 in immature neurons potentiated formation of GABA-ergic synapses (Chudotvorova et al., 2005) whereas knocking-out or knocking-down of KCC2 from cultured hippocampal neurons produced opposite effects (Medina et al., 2011). Another set of data show that the knocking-out (Li et al., 2007) or knocking-down (Gauvain et al., 2011) of the KCC2 in mature primary neuronal hippocampal cultures lead to structural modification in spines formation and decrease glutamatergic synaptic transmission. The overexpression of the KCC2 in these cultures rescued (i.e., potentiated) formation of glutamate synapses. An important task for future studies would be to determine the reason of this differential KCC2-dependent control of synaptogenesis in immature hippocampal neurons in vivo and mature hippocampal neurons in vitro.

In conclusion, we show that there is a prominent difference in neuronal network activity and formation of the functional GABA and glutamate synapses in immature CA3 hippocampal neurons of wild-type and Kcc2 deficient mice embryos. This finding provides a novel point of view on the functional importance of KCC2 during early stages of neuronal network development when GABA is still depolarizing and opens new attractive directions in study of the mechanisms of KCC2 functioning.