Properties of GABAergic Neurons Containing Calcium-Permeable Kainate and AMPA-Receptors

Calcium-permeable kainate and AMPA receptors (CP-KARs and CP-AMPARs), as well as NMDARs, play a pivotal role in plasticity and in regulating neurotransmitter release. Here we visualized in the mature hippocampal neuroglial cultures the neurons expressing CP-AMPARs and CP-KARs. These neurons were visualized by a characteristic fast sustained [Ca2+]i increase in response to the agonist of these receptors, domoic acid (DoA), and a selective agonist of GluK1-containing KARs, ATPA. Neurons from both subpopulations are GABAergic. The subpopulation of neurons expressing CP-AMPARs includes a larger percentage of calbindin-positive neurons (39.4 ± 6.0%) than the subpopulation of neurons expressing CP-KARs (14.2 ± 7.5% of CB+ neurons). In addition, we have shown for the first time that NH4Cl-induced depolarization faster induces an [Ca2+]i elevation in GABAergic neurons expressing CP-KARs and CP-AMPARs than in most glutamatergic neurons. CP-AMPARs antagonist, NASPM, increased the amplitude of the DoA-induced Ca2+ response in GABAergic neurons expressing CP-KARs, indicating that neurons expressing CP-AMPARs innervate GABAergic neurons expressing CP-KARs. We assume that CP-KARs in inhibitory neurons are involved in the mechanism of outstripping GABA release upon hyperexcitation.


Introduction
Kainate (KA) and AMPA receptors are ligand-gated channels permeable for Na + and K + . However, particular subtypes of KA and AMPA receptors are permeable for Ca 2+ . The Ca 2+ influx through these receptors can induce neurotransmitter release without the contribution of NMDARs and voltage-gated calcium channels [1].
Calcium-permeable KA receptors (CP-KARs) contain unedited GluK1 or GluK2 subunits [2,3]. The excitatory pyramidal neurons express mainly GluK2-containing KARs, whereas GluK1-containing KARs are predominantly expressed by inhibitory interneurons [4][5][6]. Most studies show that GluK1-containing KARs localize on GABAergic neurons in presynaptic terminals [3,[7][8][9]. Thus, activation of GluK1-containing kainate receptors may promote the GABA release from interneurons, leading to suppression of neuronal network activity. Contribution of GluK1-expressing GABAergic neurons in the suppression of hyperexcitation was obtained both in in vivo and in vitro experiments. Xu and coauthors demonstrated the neuroprotective effect of a selective GluK1-containing KARs agonist, ATPA, against ischemia-reperfusion-induced neuronal cell death in in vivo studies [10,11]. Furthermore, ATPA suppresses the activity of the neuronal network in cultures [12]. In turn, AMPAR/KAR agonist, DoA [13,14], increases intracellular Ca 2+ concentration ([Ca 2+ ] i ) in The tissue fragments were trypsinized for 10 min at 37 • C with constant stirring and then washed twice with cold Neurobasal-A medium to inactivate and remove trypsin. The fragments were then gently triturated with a pipette. Non-triturated tissue was removed, and the obtained suspension was centrifuged for 3 min at 2000 rpm. The sedimented cells were resuspended in Neurobasal-A medium supplemented with glutamine (0.5 mM) and B-27 (2%). Cells were seeded on round glass coverslips treated with polyethyleneimine and grown in a CO 2 -incubator (37 • C). The culture medium (2/3 of the volume) was replaced every 3 days. The density of the plated cells was 15.000 cells/sq·cm. Cell cultures at the ages of 12-14 days in vitro (DIV) were used in the experiments.

[Ca 2+ ] i Measurements. Handling of Image Data
The changes of [Ca 2+ ] i were evaluated using a fluorescent ratiometric Ca 2+ sensitive probe, Fura-2 (Molecular probes, Eugene, OR, USA), as described previously [35,37]. The cell cultures were stained for 40 min at 37 • C with freshly prepared Fura-2 AM diluted in HBSS (Hank's balanced salt solution) to the final concentration 5 µM. Then, the cells were washed with HBSS and incubated for 10-15 min to complete deestherification. To register changes of [Ca 2+ ] i , we used an inverted motorized fluorescent microscope, Axiovert 200M (Carl Zeiss Microscopy GmbH, Jena, Germany), with a high-speed monochrome CCD-cameraAxioCam HSm (Carl Zeiss Microscopy GmbH, Jena, Germany), and a high-speed light filter replacing system Ludl MAC 5000 (Ludl Electronic Products, Hawthorne, NY, USA). The reagents were added and washed using the perfusion system that provides a perfusion rate 15 mL/min. For Fura-2 excitation and registration, we used the objective Plan-Neofluar 10×/0.3, 21HE filter set (Carl Zeiss, Jena, Germany) with excitation filters BP340/30 and BP387/15, a beam splitter FT-409, and an emission filter BP510/90. Emission of Fura-2 stained cells was recorded upon excitation at the wavelengths of 340 and 387 nm. The frame rate was 1 frame per second. The recording time of one two-channel frame did not exceed 400 ms. The resulting two-channel (when Fura-2 was excited at 340 and 387 nm) time-lapse series of images were processed with ImageJ software. All experiments were carried out at 28-30 • C.

Spontaneous Synchronous Activity
Spontaneous synchronous activity (SSA) is observed throughout the brain and plays a key role in processing neuronal information, brain development, and synaptogenesis [38]. SSA is manifested in the cultures as bursts of action potentials accompanied by oscillations of [Ca 2+ ] i . It was previously shown that the number of synapses increases significantly during the first two weeks of cell growth in culture. SSA appears in the hippocampal culture a few days after the preparation and lasts for two weeks. Then, the amplitude and frequency of [Ca 2+ ] i oscillations significantly decrease due to the GABA-mediated inhibition. However, SSA can be induced in this case by GABA(A) receptor antagonists [37,39]. Oscillations of [Ca 2+ ] i during SSA are mediated by the activation of NMDA and predominantly AMPA receptors [39,40]. NBQX, an antagonist of AMPARs/KARs, abolishes the oscillations, while D-AP5, an antagonist of NMDARs, only decreases the amplitude. The SSA induced by removing GABA(A)R-mediated inhibition is rather convenient to study the interaction between cells in the cultures. The difference in agonist-induced Ca 2+ responses and parameters of SSA between the cultures of 12 and 14 days did not exceed the variation in different experiments of one day. These differences are mainly determined by the ratio of KA and AMPA receptors in neurons. The SSA in the present work was registered using [Ca 2+ ] i imaging.

Immunocytochemistry
The immunocytochemical technique was used to identify GABAergic neurons and evaluate the presence of calcium-binding proteins [36,37,41]. Since [Ca 2+ ] i measurements and the visualization of the bound antibodies were carried out using different microscopes, the marker grid was plotted on the bottom side of the coverslip with the cell culture to match the images. The recording of Fura-2 fluorescence was performed in one of the gridbordered areas. After [Ca 2+ ] i measurements, the cells in the area were photographed in the phase-contrast mode. Then, cells were fixed and stained with combinations of antibodies against glutamate decarboxylase GAD 65/67, NeuN, neuron-specific enolase (NSE), and calcium-binding proteins (parvalbumin, calbindin, calretinin).
We used freshly prepared 4% paraformaldehyde (PFA) diluted in PBS to fix cells. In the case of anti-GABA antibodies, we added 0.1% glutaraldehyde into the fixative solution to prevent GABA washing from cells during the permeabilization. The cells were incubated with PFA for 20 min and washed three times with ice-cold PBS for 5 min. To permeabilize the cells, we used 0.1% Triton X-100 solution for 15 min. Fixed cells were incubated in 10% donkey or goat serum for 30 min at room temperature to block nonspecific antibody binding sites. The cells were then incubated overnight with primary antibodies at 4 • C. We used the next primary antibodies in the present study: Neurons were distinguished from astrocytes by synchronous Ca 2+ activity and staining with mouse monoclonal anti-NeuN or anti-NSE antibodies. Astrocytes did not participate in SSA and did not respond to the KA receptor agonists and KCl (not shown).

Data Analysis
OriginLab Pro 9.1 (OriginLab, Northampton, MA, USA) and Prism GraphPad version 8.0.1 (GraphPad Software, San Diego, CA, USA) software were used for data processing, graph creation, and statistical analysis. All values are given as the mean signal of N cells, or as a typical calcium signal of most cells, or as a signal of individual neurons. Statistical analyses were performed using Student's test and Kruskal-Wallis test for group comparison. All data were obtained from at least 3 different coverslips with cells chosen from 2-3 independent passages. N-the number of cells analyzed in the individual experiment, n-the number of independent experiments. The number of cells analyzed in one experiment varied from 100 to 200.
The amplitude of [Ca 2+ ] i oscillations or the agonist-induced responses was calculated as the difference between maximum value (averaged by three time points, including the previous/next frame) of the signal (340/387 ratio) during an oscillation or response and basal level of the signal before. For calculations, as a rule, we used the averaged traces of neurons from particular groups. parvalbumin antibodies, mouse anti-calretinin antibodies, mouse anti-calbindin antibodies, donkey anti-rabbit antibodies conjugated with Alexa Fluor 647, donkey anti-mouse antibodies conjugated with Alexa Fluor 488, goat anti-rabbit antibodies conjugated with Alexa Fluor 555, goat anti-mouse antibodies conjugated with Alexa Fluor 633 (Abcam, Cambridge, UK); mouse anti-NeuN antibody (Santa Cruz Biotechnology, Dallas, TX, USA); mouse anti-NSE (neuron-specific enolase) (Bialexa, Moscow, Russia).

Identification of Neurons Expressing CP-KARs
It was previously shown that selective agonists of GluK1-containing KARs increase [Ca 2+ ] i only in certain GABAergic neurons and suppress impulse activity in other neurons in hippocampal cultures [12,15,42]. Figure 1A Figure 1C). Both Ca 2+ responses (DoA-induced in the presence of the antagonists and ATPA-induced) are generated by neurons expressing CP-KARs. We stained the cells with antibodies against GABA and NSE (neuronal marker) to demonstrate that ATPA increases [Ca 2+ ] i only in GABAergic neurons expressing CP-KARs. Figure 1C shows fluorescent images of the Fura-2-stained cells before ATPA application (15th min of the recording in Figure 1C), in the presence of ATPA (16th min), and after ATPA washing (17th min). We found that ATPA-sensitive neurons ( Figure 1C 16th min) are GABAergic since they are intensively stained with antibodies against GABA ( Figure 1C" GABA, white arrows). Thus, ATPA can be used to identify GABAergic neurons expressing CP-KA receptors.
We have previously shown that neurons expressing CP-KARs demonstrate increased excitability, possibly due to insufficient GABA(A)R-mediated inhibition [15]. We assumed that such properties allow these neurons to react more quickly to depolarization and application of agonists. To confirm this, we performed the experiments with NH 4 Cl. The ammonium ion is an endogenous toxin that causes acute or chronic hepatic encephalopathy [43,44]. It has been previously shown that ammonium (8 mM) depolarizes neurons in hippocampal cell cultures by 8-10 mV, increases basal [Ca 2+ ] i level, and induces highfrequency [Ca 2+ ] i oscillations [35,45]. Figure 1E shows that NH 4 Cl increases the basal [Ca 2+ ] i level initially only in some neurons (red and black curves), and then these cells generate [Ca 2+ ] i oscillations. The remaining neurons react to NH 4 Cl with synchronous [Ca 2+ ] i oscillations appearing after a delay (67 ± 15s) (blue and purple curves) without preliminary [Ca 2+ ] i rise. ATPA application before the experiment (application not shown) revealed that 42 ± 11% of fast-responding neurons were ATPA-sensitive ( Figure 1E, black curves). Even low doses of ATPA increases [Ca 2+ ] i in these neurons, indicating their greater excitability ( Figure 1F, black curves). Thus, GABAergic neurons expressing GluK1-containing CP-KARs react to NH 4 Cl-induced depolarization faster than most glutamatergic neurons in mature hippocampal cell culture. Therefore, NH 4 Cl-induced Ca 2+ influx may stimulate GABA release, which suppresses the excitation of glutamatergic neurons. ( Figure 1F, black curves). Thus, GABAergic neurons expressing GluK1-containing CP-KARs react to NH4Cl-induced depolarization faster than most glutamatergic neurons in mature hippocampal cell culture. Therefore, NH4Cl-induced Ca 2+ influx may stimulate GABA release, which suppresses the excitation of glutamatergic neurons.

Identification of Neurons Containing CP-AMPARs
To visualize neurons expressing CP-AMPARs and CP-KARs in the same experiment, we used low doses of DoA (300 nM). DoA activates KARs and AMPARs [13,14,46] and causes the influx of Ca 2+ into neurons through CP-KA and CP-AMPA receptors. The experiment was performed in the presence of bicuculline to abolish GABA(A)R-mediated inhibition and induce SSA [12,37,40,47,48]. Applications of DoA in the absence and the presence of NASPM (a CP-AMPAR antagonist) were made to distinguish neurons expressing CP-KARs from neurons containing CP-AMPARs (Figure 2A). A short-term application of 300 nM DoA causes a rapid, sustained increase in basal [Ca 2+ ] i in some neurons (red and black curves in Figure 2D). In all other neurons, DoA increases the frequency of [Ca 2+ ] i oscillations (blue curves in Figure 2D Figure 2D, red curves). Therefore, the second subpopulation can be attributed to neurons expressing CP-AMPARs. Figure 2B shows that 15 ± 3.6% of neurons respond to DoA with a sustained increase of basal [Ca 2+ ] i . NASPM-sensitive neurons accounted for 71.6 ± 5.5% of DoA-responding neurons ( Figure 2C). The number of NASPM-insensitive neurons coincided with the number of neurons that responded to ATPA, so they can be attributed to neurons containing GluK1 subunit.

Neurons Expressing CP-AMPARs Are GABAergic
To prove that all neurons responding to DoA with a sustained [Ca 2+ ]i increase (red and black curves in Figure 2D) are GABAergic, we stained the cultures with antibodies against GAD 65/67 (a marker of GABAergic neurons) and the neuronal marker, NeuN ( Figure 3C). Before immunostaining, we identified these neurons by the high level of [Ca 2+ ]i after 100 s DoA exposure ( Figure 3B, the bottom image). activated kainate receptors are insensitive to bicuculline-induced postsynaptic membrane depolarization [42], implying the absence of GluK1-containing CP-KARs in the postsynaptic membrane of these neurons. Figures 1A and 2C show that NASPM-insensitive neurons responding to DoA application with a sustained [Ca 2+ ] i increase also respond to ATPA, indicating the presence of GluK1-containing CP-KARs. Thus, the activation of DoA-sensitive neurons probably stimulates GABA release, which suppresses calcium response in glutamatergic neurons in the absence of the GABA(A)R antagonist. Thus, DoA-and ATPA-induced Ca 2+ responses make it possible to visualize GABAergic neurons expressing CP-KARs and CP-AMPARs. The CP-KARs and CP-AMPARs are expressed in different neurons and in the different terminals: CP-AMPARs in postsynaptic terminals and CP-KARs in presynaptic.

Neurons Expressing CP-AMPARs Are GABAergic
To prove that all neurons responding to DoA with a sustained [Ca 2+ ] i increase (red and black curves in Figure 2D) are GABAergic, we stained the cultures with antibodies against GAD 65/67 (a marker of GABAergic neurons) and the neuronal marker, NeuN ( Figure 3C). Before immunostaining, we identified these neurons by the high level of [Ca 2+ ] i after 100 s DoA exposure ( Figure 3B, the bottom image).
Immunostaining revealed that the neurons responding to DoA with a sustained [Ca 2+ ] i increase mainly belong to the GABAergic ( Figure 3C,F). In turn, the neurons that reacted to DoA with the delayed increase in SSA frequency (blue and green curves in Figure 2D,E and Figure 3A) were not stained with antibodies against GAD65/67 and can be attributed to the glutamatergic neurons. Thus, we showed that both subpopulations of neurons responding to DoA and NH 4 Cl with a sustained [Ca 2+ ] i increase are GABAergic. Neurons of one subpopulation express CP-KARs, and the other one-CP-AMPARs. In the hippocampal culture, 29 ± 4% of neurons were stained with anti-GAD65/67 antibodies ( Figure 3D); DoA-sensitive GABAergic neurons account for approximately 39 ± 2% of GAD 65/67-positive cells ( Figure 3E). Among them, 32.6 ± 6% of neurons express CP-KARs ( Figure 2C). The fraction of GAD65/67-negative cells does not practically include neurons that express CP-KARs and CP-AMPARs ( Figure 3F).

Calcium-Binding Proteins in Neurons Expressing CP-AMPARs and CP-KARs
GABAergic neurons can express various calcium-binding proteins (CBPs) [49,50]. Interestingly, some CBPs are exclusively expressed by GABAergic neurons and are considered their specific markers [49,51]. To determine the presence of CBPs in GABAergic neurons expressing CP-KARs or CP-AMRARs, we used double staining with antibodies against parvalbumin (PV) and GAD 65/67 ( Figure 4A Neurons expressing CP-AMPARs and CP-KARs were identified by the sensitivity to NASPM and by the presence of ATPA-induced Ca 2+ -response described in Sections 3.1 and 3.2 (Figures 1 and 2). Figure 4B,C demonstrate the percentage of the PV + , CB + , and CR + GAD 65/67-positive neurons expressing CP-KARs ( Figure 4B) and CP-AMPARs ( Figure 4C). Figures show that both groups of neurons include a significant number of parvalbumin-containing cells (62.6 ± 11.5% for CP-AMPARs-containing and 64.8 ± 11.0% for CP-KARs-containing) and a small number of calretinin-containing cells (24.6 ± 10.1% for CP-AMPARs-containing and 32.4 ± 5.3% for CP-KARs-containing). However, the subpopulation of neurons expressing CP-AMPARs includes more CB + neurons (39.4 ± 6.0%) than the subpopulation of neurons expressing CP-KARs (14.2 ± 7.5% of CB + neurons). Thus, the presence of calbindin in neurons expressing CP-AMPARs may partially explain the slower calcium response of these neurons to DoA.  Immunostaining revealed that the neurons responding to DoA with a sustained [Ca 2+ ]i increase mainly belong to the GABAergic ( Figure 3C,F). In turn, the neurons that reacted to DoA with the delayed increase in SSA frequency (blue and green curves in Figure 3A, and Figure 2D,E) were not stained with antibodies against GAD65/67 and can be attributed to the glutamatergic neurons. Thus, we showed that both subpopulations of  GABAergic neurons can express various calcium-binding proteins (CBPs) [49,50]. Interestingly, some CBPs are exclusively expressed by GABAergic neurons and are considered their specific markers [49,51]. To determine the presence of CBPs in GABAergic neurons expressing CP-KARs or CP-AMRARs, we used double staining with antibodies against parvalbumin (PV) and GAD 65/67 ( Figure 4A

GABAergic Neurons Expressing CP-AMPARs Inhibit GABAergic Neurons Expressing CP-KARs
Fluorescent Ca 2+ imaging allows to record changes in the activity of target (innervated) neurons in response to changes in the activity of GABAergic neurons [15,42]. To identify neurons innervated by GABAergic neurons expressing CP-KARs and CP-AMRARs, we analyzed the amplitude and frequency of SSA and DoA-induced calcium responses in control and in the presence of NASPM ( Figure 5A,D) in four neuronal subpopulations from the experiment shown in Figure 2A. The percentage of neurons in each subpopulation is demonstrated in Figure 5F. As can be seen ( Figure 5E, DoA increases basal [Ca 2+ ] i in 28 neurons (green and red markers) out of 125 in a view field in control. In the presence of NASPM, DoA suppresses the signal in 17 neurons and increases in 9 neurons out of these 28.
In the presence of NASPM, the amplitude of [Ca 2+ ] i oscillations after DoA washout did not change in CP-KARs-expressing GABAergic neurons ( Figure 5A, CP-KAR). This finding confirms the lack of CP-AMPARs in these neurons. However, a significant increase in the amplitude of DoA-induced Ca 2+ response in neurons of this subpopulation in the presence of NASPM indicates that these neurons are innervated by GABAergic neurons containing CP-AMPARs. The amplitude of the DoA-induced Ca 2+ response in the presence of NASPM did not increase in any other neurons. It seems that NASPM inhibits Ca 2+influx and Ca 2+ -dependent GABA release in interneurons expressing CP-AMPARs, thus abolishing GABA-mediated inhibition of neurons expressing CP-KARs. The amplitude of  The amplitude of [Ca 2+ ] i oscillations in one subpopulation of glutamatergic neurons (Glut-1) depended on the presence of NASPM and did not depend in the second one (Glut-2) ( Figure 5C,D). The amplitude of SSA decreased in glutamatergic neurons (Glut-1) in the presence of NASPM. Since glutamatergic neurons do not virtually express CP-KARS and CP-AMPARs ( Figure 3F), we can assume that Glut-1 neurons are innervated by GABAergic neurons expressing CP-KARs, which activity significantly increases at this time ( Figure 5A, CP-KAR). The SSA amplitude does not change in glutamatergic neurons (Glut-2) in the presence of NASPM. The amplitude of DoA-induced calcium oscillations also did not change in the presence of NASPM, indicating that this subpopulation of neurons is not innervated by GABAergic neurons expressing CP-AMRARs or CP-KARs. Thus, this experiment suggests that GABAergic neurons expressing CP-AMPARs innervate GABAergic neurons expressing CP-KARs, which, in turn, control the numerous subpopulation of glutamatergic neurons.

Discussion
Two subpopulations of GABAergic neurons expressing CP-KARs and CP-AMPARs were detected in rat hippocampal cell cultures on 12-14 days in vitro (DIV). The neurons expressing CP-KARs were identified by the [Ca 2+ ] i increase in response to ATPA. ATPAinduced Ca 2+ response was insensitive to antagonists of NMDA and AMPA receptors, indicating that ATPA-induced depolarization is not enough to activate NMDARs. Neurons expressing CP-AMPARs were identified by the sensitivity of DoA-induced [Ca 2+ ] i increase to the antagonist of CP-AMPA receptors, NASPM. We found in the present work that CP-KARs and CP-AMRARs are localized mainly on various subpopulations of GABAergic neurons. The results are consistent and complement the data, according to which GluK1containing KA receptors are mainly expressed in a certain subpopulation of GABAergic neurons in the hippocampus [4][5][6]. It was also shown that kainate causes rapid Ca 2+ influx and [Ca 2+ ] i increase in distinct subpopulations of GABAergic neurons [52]. Agonists of CP-KA and CP-AMPA receptors selectively increase [Ca 2+ ] i in these neurons, leading to the GABA release and inhibition of other different neurons in the network [31]. This inhibition is largely abolished by the antagonist of GABA(A) receptors, bicuculline [15].
Unlike ATPA, DoA induces a sustained increase of basal [Ca 2+ ] i in both CP-KARsand CP-AMPARs-expressing GABAergic neurons. In turn, [Ca 2+ ] i oscillations in glutamatergic neurons (Glut-1 and Glut-2) appeared with a delay and were not followed by a sustained increase in basal [Ca 2+ ] i . Early DoA-induced [Ca 2+ ] i rise is not accompanied by neuronal excitation, and [Ca 2+ ] i oscillations appeared in all neurons only after 60 s of DoA exposure. Considering the existence of voltage thresholds for the spreading of depolarizing stimulus, it may be suggested that the currents mediated by CP-KARs and CP-AMPARs are insufficient to excite GABAergic neurons but sufficient to trigger the GABA release. Slower Ca 2+ removal in these neurons may promote faster Ca 2+ accumulation in the cytosol, potentiating GABA release.
The massive GABA release due to the faster response of GABAergic neurons expressing CP-AMPARs and CP-KARs may explain the delayed response in Glut-1 and Glut-2 neurons. This assumption also explains the delay of response in most neurons after NH 4 Cl application ( Figure 1E). As previously shown [16], DoA-sensitive neurons are more excitable than other neurons, probably due to insufficient GABA(A)R-mediated inhibition. This feature probably allows GABAergic neurons to respond to an excitatory stimulus earlier compared to other neurons (glutamatergic) and release GABA, thus suppressing the activity of the innervated neurons. In addition, the higher excitability of GABAergic neurons may be explained by the altered KCC2 expression, affecting Cl − gradient and the response of neurons to GABA [53,54].
It is known that GABAergic neurons are subdivided into several subtypes that differ in the expression of calcium-binding proteins [49,50,55]. The role of these proteins is actively discussed [56,57]. Calcium-binding proteins protect GABAergic neurons against a global [Ca 2+ ] i increase under oxygen-glucose deprivation [41]. Here we showed the correlation between the rate of [Ca 2+ ] i increase and the presence of CBPs in neurons. We found that a subpopulation of neurons expressing CP-AMPARs includes a larger percentage of CB + neurons (39.4 ± 6.0%) than the subpopulation of neurons expressing CP-KARs (14.2 ± 7.5% of CB + neurons). The presence of calbindin in neurons expressing CP-AMPARs may explain the slower calcium response of these neurons to DoA [34]. Thus, a subpopulation of GABAergic neurons expressing CP-AMPARs may include a subtype of neurons expressing calbindin. Fast-binding Ca 2+ proteins, such as CB and CR, decrease the rate of [Ca 2+ ] i rise and limit the amplitude of the Ca 2+ signal, thus slowing or preventing GABA release at low frequencies of [Ca 2+ ] i oscillations. However, when the cell is hyperexcited, the average rate of Ca 2+ entry into cytosol exceeds the pumping rate. Repletion of the intracellular Ca 2+ buffers will increase the amplitude of [Ca 2+ ] i changes and, accordingly, increase the intensity of neurotransmission. This mechanism can be used to suppress the activity of target neurons under hyperexcitation. Calcium-binding proteins can also protect neurons from damage under oxygen and oxygen-glucose deprivation [41].
We analyzed the [Ca 2+ ] i changes of hundreds of neurons to establish interaction between neuronal subpopulations. We showed that inhibition of CP-AMRARs by NASPM causes overexcitation of GABAergic neurons expressing CP-KARs ( Figure 5A), pointing out that GABAergic neurons expressing CP-AMPARs inhibit the activity of GABAergic neurons expressing CP-KARs. This mechanism may be considered negative feedback, which aims to suppress the increased activity of GABAergic neurons expressing CP-KARs.
A decrease in the activity of some glutamatergic neurons in the presence of NASPM correlates with increased activity of neurons expressing CP-KARs ( Figure 5A,C; Glut-1), thus indicating the innervation of these glutamatergic neurons by inhibitory neurons expressing CP-KARs. We also found a subpopulation of glutamatergic neurons, which do not respond to the antagonist of CP-AMPARs ( Figure 5D, Glut-2). Probably, these neurons do not contain CP-AMPARs and are not innervated by NASPM-sensitive neurons. This finding agrees with the data of Wondolowski and colleagues [58], showing that interneurons expressing KARs innervate only a specific subpopulation of neurons. The inhibition of cholecystokinin-releasing interneurons innervating pyramidal cells by GABAergic neurons with presynaptic KARs also has been shown [59]. Our data suggest that GABAergic neurons expressing CP-KARs and CP-AMPARs can suppress hyperexcitation of other neurons due to rapid, sustained increase of [Ca 2+ ] i followed by GABA release. Thus, during the development of neurons in culture for 14 days, the neuronal network self-organizes, and interactions between neurons expressing CP-AMPARs, and CP-KARs are formed. There is no direct evidence for the presence of such mechanisms in vivo, although indirect experiments indicate this possibility [24][25][26][27][28][29]. The earlier response of GABAergic neurons expressing GluK1-containing KARs to depolarization may be considered as an inhibition mechanism in neuronal networks.

Conclusions
We showed a new mechanism of the firing activity regulation in hippocampal cell culture by GABAergic neurons expressing CP-KARs and CP-AMPARs. Two different subpopulations of GABAergic neurons were identified. One subpopulation contains CP-KARs, and the other one contains CP-AMPARs. CP-KARs and CP-AMPARs agonists cause a rapid [Ca 2+ ] i increase followed by GABA release in these neurons, thus suppressing the activity of other neurons. In addition, these GABAergic neurons respond faster than other neurons to NH 4 Cl-induced depolarization. The innervation of GABAergic neurons expressing CP-KARs by GABAergic neurons expressing CP-AMPARs explains the excitatory effect of agonists and the neuroprotective effect of antagonists of CP-AMPARs.
The obtained data suggest that GABAergic neurons expressing CP-KARs and CP-AMPARs can inhibit hyperexcitation of other neurons due to faster reaction to an excitatory stimulus leading to GABA secretion.