Physiological Role of ATPase for GABAA Receptor Resensitization

γ-Aminobutyric acid type A receptors (GABAARs) mediate primarily inhibitory synaptic transmission in the central nervous system. Following fast-paced activation, which provides the selective flow of mainly chloride (Cl−) and less bicarbonate (HCO3−) ions via the pore, these receptors undergo desensitization that is paradoxically prevented by the process of their recovery, referred to as resensitization. To clarify the mechanism of resensitization, we used the cortical synaptoneurosomes from the rat brain and HEK 293FT cells. Here, we describe the effect of γ-phosphate analogues (γPAs) that mimic various states of ATP hydrolysis on GABAAR-mediated Cl− and HCO3− fluxes in response to the first and repeated application of the agonist. We found that depending on the presence of bicarbonate, opened and desensitized states of the wild or chimeric GABAARs had different sensitivities to γPAs. This study presents the evidence that recovery of neuronal Cl− and HCO3− concentrations after desensitization is accompanied by a change in the intracellular ATP concentration via ATPase performance. The transition between the desensitization and resensitization states was linked to changes in both conformation and phosphorylation. In addition, the chimeric β3 isoform did not exhibit the desensitization of the GABAAR-mediated Cl− influx but only the resensitization. These observations lend a new physiological significance to the β3 subunit in the manifestation of GABAAR resensitization.


Introduction
As members of the pentameric ligand-gated ion channels family (pLGIC), γ-Aminobutyric acid type A receptors (GABA A Rs) are mainly involved in inhibition but also in excitation of the central nervous system of vertebrates under certain circumstances [1,2]. Upon mediator binding, GABA A Rs quickly open their transmembrane pore to enable the passive flow of chloride (Cl − ) and less bicarbonate (HCO 3 − ) ions via the plasma membrane. However, following activation, GABA A Rs undergo desensitization, which involves inchmeal entry into the long-lived closed state refractory to excessive activation [3][4][5][6]. Although the causes and the roles of receptor desensitization continue to be debated, they potentially include the reduction in responses during high-frequency neurotransmitter release and the prolongation of synaptic currents [7,8]. In addition, the GABA A R desensitization is extremely related to the slowdown of their deactivation, called resensitization [9,10]. The linkage between desensitization and resensitization of GABA A R-mediated currents is demonstrated in a response to a prolongation [2,5] or reapplication of GABA [3,11]. However, despite the critical importance of GABA A R resensitization in maintaining their responsiveness to subsequent activation, it is unclear which molecular events trigger this long-term receptor alteration.
The regulation of desensitization and resensitization of GABA A Rs is an important mechanism controlled at the receptor level and within the signaling pathway, but regulation of the receptors assumes great significance in understanding the withdrawal of neurological disorders [5]. In addition, the regulation of GABA A R neurotransmission plays an exceptional role in psychiatric disabilities and addiction [12,13]. Many studies of desensitization have focused on changes to the receptor following massive activation with agonists or allosteric ligands that affect the occupancy of this state during synaptic inhibition [9,14]. Such changes can include phosphorylation of subunits by kinases [5], expression, clustering [15], and receptor pharmacology [16]. Additionally, some structural works have shown that the differences between the agonist-bound open (conducting) state and desensitization (hindering) state are conformational changes occurring at the "internal face" of the receptor. Recent functional and structural studies provide compelling evidence for a "dual-gate" model, in which the transmembrane domain (TMD) of pLGICs contains both an activation gate, located in the upper half of the channel, and a de-gate, at the intracellular end of the channel [17,18].
Preliminary studies have shown that the GABA A R-mediated current is significantly elevated by inhibitors of phosphatases [19,20] or non-hydrolysable ATP analogues [21], indicating the involvement of the processes of phosphorylation and dephosphorylation in regulating GABA A R functional activity [22]. It has also been demonstrated during intensive agonist action that the diminution in intracellular adenosine triphosphate concentration ([ATP] i ) is the main reason underpinning the functional rundown of GABA A R function, eventually paving the way for neuronal excitation [23][24][25][26]. However, the molecular determinants that provide the ATP depletion within the GABA A R rundown phenomenon remain poorly understood. Recently, we reported that the GABA A R β3 subunit possesses Cl − and HCO 3 − ATPase activity, which provides the anion homeostasis and is involved in neurological disorders [27,28]. Given that such an ATPase is specifically regulated by GABAergic agonists or antagonists [27] and the ATP-hydrolyzing center is localized at the intracellular side of the channel [29], we set out to test the hypothesis about its participation in the processes of the desensitization and resensitization. Specifically, our focus was on γ-phosphate-modified ATP analogues (γPAs) that are often used to mimic various important states of ATP hydrolysis: non-hydrolysable ATP analogue adenylyl-imidodiphosphate (AMP-PNP)-the pre-hydrolytic state, and vanadate (VO 4 3− ) and fluoride metal (AlF x )the transition state [30]. We found that activation (open) and desensitization states of GABA A R-mediated Cl − and HCO 3 − flows in the synaptoneurosomes (SNs) of the cortical neurons of the rat brain were different in terms of the sensitivity to γPAs. The transition between the desensitization and resensitization states was correlated with the changing intracellular concentration of ATP ([ATP] i ) determined by HCO 3 − . The β3 subunit was established as being essential for the process of GABA A R resensitization by ATPase performance. By using the thiol-modifying agent n-ethylmaleimide (NEM), we provided evidence of differences in the conformational properties of activation and desensitization states of GABA A Rs. The important role of conformational rearrangements and the resulting changes in the processes of phosphorylation and dephosphorylation was confirmed by using a mutant receptor that substantially alters the receptor's ability to access the transition from a desensitized state to a resensitized state.

Reapplication of GABA Causes Desensitization of GABA A R-Mediated Cl − Influx
The preliminary studies have shown that the GABA A Rs of the SNs provide the Cl − influx via plasma membranes [31][32][33]. Earlier, we showed that GABA (1-100 µM) increased the ATPase activity of the ternary (α2β3γ2) GABA A R subtype with maximum effect at 100 µM [27]. To explore the GABA A R-mediated Cl − influx, the SNs initially were loaded with dye for Cl − (MQAE) and then exposed to 10, 50, or 100 µM of GABA. The SNs in the HCO 3 − -free medium showed a Cl − influx in response to the application of GABA with a maximum peak of fluorescence changes of 2.6 ± 0.3%, 4.4 ± 0.3%, or 6.7 ± 0.2%, respectively. Therefore, as illustrated in Figure 1A, we used GABA at a concentration of 100 µM. Some studies have shown that the repetitive application of the same agonist concentration can induce the desensitization of the GABA-evoked Cl − current [11]. We applied second applications of the agonist to assess the nature of agonist-dependent desensitization and resensitization of GABA A R. However, reapplication of GABA (100 µM) revealed apparent desensitization of the GABA A R-mediated Cl − influx that is similar to literature data. Specifically, the studies showed that in rat central neurons, the GABA A R-mediated Cl − current leads to substantial desensitization with the repetitive application of GABA [11]. creased the ATPase activity of the ternary (α2β3γ2) GABAAR subtype with maximum effect at 100 μM [27]. To explore the GABAAR-mediated Cl − influx, the SNs initially were loaded with dye for Cl − (MQAE) and then exposed to 10, 50, or 100 μM of GABA. The SNs in the HCO3 − -free medium showed a Cl − influx in response to the application of GABA with a maximum peak of fluorescence changes of 2.6 ± 0.3%, 4.4 ± 0.3%, or 6.7 ± 0.2%, respectively. Therefore, as illustrated in Figure 1A, we used GABA at a concentration of 100 μM. Some studies have shown that the repetitive application of the same agonist concentration can induce the desensitization of the GABA-evoked Cl − current [11]. We applied second applications of the agonist to assess the nature of agonist-dependent desensitization and resensitization of GABAAR. However, reapplication of GABA (100 μM) revealed apparent desensitization of the GABAAR-mediated Cl − influx that is similar to literature data. Specifically, the studies showed that in rat central neurons, the GABAARmediated Cl − current leads to substantial desensitization with the repetitive application of GABA [11].  In exploring and establishing the role of the ATP-hydrolyzing system in the desensitization and resensitization, we studied the effect of VO4 3− (20 μM), AlFx (20 μM), or AMP- In exploring and establishing the role of the ATP-hydrolyzing system in the desensitization and resensitization, we studied the effect of VO 4 3− (20 µM), AlF x (20 µM), or AMP-PNP (2 mM) on the GABA A R-mediated Cl − influx. All test γ-PAs caused an increase in the GABA-mediated Cl − influx into neurons by approximately two times at first and had no effect on the second application of the mediator ( Figure 1A). Bicuculline completely suppressed the first peak of the GABA-mediated fluorescence changes, indicating the receptor-dependent method of mediator action ( Figure 1B). By contrast, at the second application of the mediator, bicuculline significantly activated the Cl − influx in the SNs with a maximum peak of fluorescence changes of 11.0 ± 0.5% (n = 5) ( Figure 1C).
It has been shown that synaptic sulfhydryl groups of ionic channels and transporters are targets for electrophiles [34,35]. Specifically, NEM has been used as an agent to regulate the cation-chloride cotransporter (CCC) function by modulating their phosphorylation [35]. In addition, NEM (300 µM) causes an increase in the frequency of the GABA A R-mediated postsynaptic currents [36,37] and eliminates the depolarization-or post-burst-induced suppression of GABA A R-mediated inhibition in CA1 pyramidal cells [38,39]. Here, NEM (300 µM) activated both at first and at the reapplication of GABA with a maximum peak of fluorescence changes of 17.4 ± 1.9% (n = 5) and 9.8 ± 1.9% (n = 5), respectively ( Figure 1D,E). NEM effect did not change significantly amid the presence of VO 4 3− (20 µM) in the first case and reapplication of the agonist ( Figure 1D,E).

Bicarbonate Determines a Transition from Desensitization to Resensitization State of
The preliminary studies have shown that HCO 3 − increased the GABA A R-mediated Cl − influx in the neurons [40]. In our study, the SNs in the presence of HCO 3 − (25 mM) showed increased Cl − influx in response to the first application of GABA (100 µM) with a maximum peak of fluorescence changes of 13.7 ± 2.1% (n = 5) ( Figure 2A). Reapplication of GABA (100 µM) revealed a decrease in the GABA A R-mediated Cl − influx by two times with a maximum peak of fluorescence changes of 6.5 ± 1.0% (n = 5). Previously, it was shown that the effect of ATP and γPAs on GABA A R-mediated Cl − transport in the neurons depends on the presence of HCO 3 − in the experimental medium [22][23][24]. To probe whether the HCO 3 − can regulate the GABA A R function, we studied the effect of γPAs on the GABA A R-mediated Cl − influx in the presence of a physiological concentration of HCO 3 − (25 mM) ( Figure 2A). Here, γPAs eliminated the GABA-mediated Cl − influx into neurons at the first application and activated it approximately two times on reapplication of the agonist (Figure 2A). Bicuculline completely suppressed the first peak of the GABA-mediated fluorescence changes, confirming the receptor-dependent method of mediator action ( Figure 2B). In contrast, at reapplication of the mediator, bicuculline had no significant influence (7.5 ± 0.8%, n = 6) on the Cl − influx in the SNs ( Figure 2C). Similar to bicuculline's effect, NEM (300 µM) inhibited (1.3 ± 1.2%, n = 6) on the first application and had no effect (6.5 ± 1.0%, n = 6) on repeated application of GABA ( Figure 2D). Meanwhile, in the presence of VO 4 3− (20 µM) in the experimental medium, the NEM effect was eliminated in the first case ( Figure 2D) and unchanged at reapplication of GABA ( Figure 2E).

ATP-Dependent Recovery of GABA A R-Mediated pH i Changes
The rat hippocampal pyramidal neurons show an instantaneous net influx of Cl − and efflux of HCO 3 − in response to GABA (500 µM) application with a subsequent steadystate decrease in intracellular pH (pH i ) of 0.2 to 0.5 [41,42]. To explore the GABA A Rmediated HCO 3 − outflux, the SNs initially were loaded with dye for pH i (BCECF) before being exposed to GABA (100 µM). Since at reapplication of the same concentration of GABA (100 µM) [43,44], the GABA A Rs are permeable to HCO 3 − , we studied the change in the pH i in the SNs after the first and repeated application of GABA (100 µM) in the presence of HCO 3 − (25 mM). Here, similar results were observed, where SNs quickly demonstrated a pH i decrease for both the first and repetitive application of the agonist with a maximum peak of fluorescence changes of 15.5 ± 1.5% (n = 6) and 12.0 ± 1.0% (n = 6), respectively ( Figure 3B,C). These data showed the absence of desensitization and only appeared in the process of resensitization. To further elucidate the cause of pH i changes by GABA application, we added bicuculline to the experimental medium. Bicuculline (50 µM) inhibited the GABA-mediated HCO 3 − outflux to first and repeated application of the agonist ( Figure 3D,E). We added the γPAs in the experimental medium to assess the role of ATPase in the recovery of GABA A R-mediated HCO 3 − outflux. All γPAs eliminated the GABA-mediated pH i changes in the first and repeated application of the agonist ( Figure 3A-C).
When exposed to NEM (300 µM), the GABA-mediated pH i changes were significantly eliminated for both the first and repeat application of GABA ( Figure 3F,G). However, the NEM effect did not appear in the presence of VO 4 3− (20 µM), thus indicating the energydependent method of action ( Figure 3F,G). These data suggest that NEM and HCO 3 − have similar binding sites that are close to the ATP-hydrolyzing center. In order to ascertain this hypothesis, we examined the influence of HCO 3 − and NEM on [ATP] i . In the control samples, the concentration of [ATP] i was 3.0 nmol/mg protein and the addition of HCO 3 − (25 mM) induced a decrease in [ATP] i by 29.2 ± 1.0% (n = 6) ( Figure 3H). Moreover, as shown in Figure 3I, the addition of NEM in the experience medium also resulted in a decline in [ATP] i by 37.6 ± 1.2% (n = 5). However, the influence of NEM on [ATP] i did not appear in the presence of VO 4 3− (20 µM). When exposed to NEM (300 μM), the GABA-mediated pHi changes were significantly eliminated for both the first and repeat application of GABA ( Figure 3F,G). However, the NEM effect did not appear in the presence of VO4 3− (20 μM), thus indicating the energy-dependent method of action ( Figure 3F,G). These data suggest that NEM and HCO3 − have similar binding sites that are close to the ATP-hydrolyzing center. In order to ascertain this hypothesis, we examined the influence of HCO3 − and NEM on [ATP]i. In the control samples, the concentration of [ATP]i was 3.0 nmol/mg protein and the addition of HCO3 − (25 mM) induced a decrease in [ATP]i by 29.2 ± 1.0% (n = 6) ( Figure 3H). Moreover, Tukey's test: F (9,6) = 11.0 or F (9,6) = 0,005, p = 0.05, n = 5). All data in the bar graphs in this figure and those proceeding it are presented as mean values +/− SEM; * p < 0.05.

β3 Subunit Is Responsible for GABA A R Resensitization in HEK 293FT Cells
Previously, we have shown that only the GABA A R β3 subunit, in contrast to α2 or γ2 subunits, participates in the GABA-mediated or ATP-dependent Cl − transport [27]. To examine whether the β3 subunit is also involved in the HCO 3 − transport, we studied the properties of the GABA A R-mediated HCO 3 − outflux in the HEK 293FT cells expressing the homomeric GABA A R β3 isoform ( Figure 4A). HEK 293FT cells expressing the GABA A R β3 isoform showed one band in the VLPs with a molecular weight of approximately~54 kDa that was bound to antibodies against the GABA A R β3 subunits ( Figure 4B). The HEK 293FT cells in the presence of HCO 3 − (25 mM) in the experience medium showed a rapid pH i decrease in the both first and repetitive applications of the agonist with a maximum peak of fluorescence changes of 10.2 ± 0.4% (n = 5) and 9.7 ± 0.5% (n = 5) ( Figure 4C,D), respectively, and with a subsequent rapid (~40 s) recovery of [HCO 3 − ] i ( Figure 4A). γPAs completely eliminated the GABA-mediated fluorescence changes in the first application and at reapplication of GABA (100 µM) ( Figure 4A,C,D). Bicuculline (50 µM) completely eliminated the fluorescence changes, recorded in response to first and reapplication of GABA (100 µM) in an experimental medium ( Figure 4E,F).
subunits, participates in the GABA-mediated or ATP-dependent Cl − transport [27]. To examine whether the β3 subunit is also involved in the HCO3 − transport, we studied the properties of the GABAAR-mediated HCO3 − outflux in the HEK 293FT cells expressing the homomeric GABAAR β3 isoform. HEK 293FT cells expressing the GABAAR β3 isoform showed one band in the VLPs with a molecular weight of approximately ~54 kDa that was bound to antibodies against the GABAAR β3 subunits ( Figure 4B). The HEK 293FT cells in the presence of HCO3 − (25 mM) in the experience medium showed a rapid pHi decrease in the both first and repetitive applications of the agonist with a maximum peak of fluorescence changes of 10.2 ± 0.4% (n = 5) and 9.7 ± 0.5% (n = 5) ( Figure 4C,D), respectively, and with a subsequent rapid (~40 s) recovery of [HCO3 − ]i ( Figure 4A). γPAs completely eliminated the GABA-mediated fluorescence changes in the first application and at reapplication of GABA (100 μM) ( Figure 4A,C,D). Bicuculline (50 μM) completely eliminated the fluorescence changes, recorded in response to first and reapplication of GABA (100 μM) in an experimental medium ( Figure 4E,F).  NEM (300 µM) activated the HCO 3 − outflux at the first and repeated application of GABA with a maximum peak of fluorescence changes of 19.2 ± 1.1% (n = 5) and 13.5 ± 1.4% (n = 5), respectively ( Figure 4G,H). The NEM effect did not appear in the presence of VO 4 3− (20 µM) assuming the involvement of the ATP-hydrolyzing system in the process of [HCO 3 − ] i recovery. To assess the role of ATPase in the consuming of [ATP] i , we studied the effect of GABA on the Cl − , HCO 3 − ATPase activity. As shown in Figure 4I, the control GABA A R β3 isoform showed Cl − , HCO 3 − ATPase activity of 0.27 ± 0.2 µmol −1 P i min −1 mg −1 . The addition of GABA (100 µM) induced increased enzyme activity by one and a half times.
To examine the contribution of the β3 subunit in the processes of resensitization, we selected a mutation nearby the desensitization-gate, which was previously found to specifically alter the desensitization kinetics and amplitude [10,17,18]. Specifically, a single cysteine residue (C313) in the M3 domain was conserved in all GABA A R β subunits and demonstrated a specific response to modulation by oxidizing agents [45]. To test whether this cysteine residue formed at least in part the molecular basis for NEM modulation, this residue was mutated to alanine in the β3 subunit (C313A). HEK 293FT cells expressing the mutant GABA A R β3 isoform showed one band in the VLPs with a molecular weight of approximately~54 kDa that was bound to antibodies against the GABA A R β3 subunits, respectively ( Figure 4J). In contrast to the control receptor isoform, the homomeric mutant GABA A R β3 (C313A) isoform displayed the GABA-mediated recovery of pH i (8.9 ± 0.7%) in the first case and at the second application of GABA, showing desensitization ( Figure 4K). The mutant GABA A R β3 isoform also possessed Cl − , HCO 3 − ATPase activity (0.23 ± 0.01 µmol −1 P i min −1 mg −1 ), and an exposition with GABA (100 µM) only decreased the enzyme activity ( Figure 4L). Meanwhile, in the HEK 293FT cells expressing the homomeric GABA A R β3 isoform, GABA (100 µM) in both the first and repetitive application induced Cl − influx in cells, but it was not statistically significant ( Figure 4M,N). The maximal peak of fluorescence changes for the first and repeated application was 2.9 ± 0.8 (n = 5) and 2.3 ± 0.9 (n = 5), respectively, and that had few changes in the presence of VO 4 3− (20 µM) or AMP-PNP (2 mM).

The Reconstituted Cl − , HCO 3 − ATPase Responsible for GABA A R Resensitization
We followed-up on our results showing that the β3 subunit alone sufficed to show ATPase activity by transfecting HEK 293FT cells with plasmid vectors containing GABA A R cDNA of the β3 subunit bound to a His-tag fusion protein [27]. These studies showed that reconstituted GABA A Rs can participate in both the influx and efflux of Cl − . We purified and reconstituted the enzyme and investigated the ATP-dependent Cl − -transport in proteoliposomes containing the embedded protein and the fluorescent dye MQAE. To understand how the β3 subunit might affect GABA A R resensitization, the proteoliposomes after addition of 2 mM Mg 2+ -ATP were exposed with the agonist.
As shown in Figure 5A    The first addition of GABA (100 µM) to the incubation medium caused a short-term (~1 s) efflux of Cl − from the proteoliposomes with a maximal peak of fluorescence change of 18.0 ± 1.3% (n = 4) that was quickly restored and reached a plateau during~60 s of incubation ( Figure 5A). Repetitive application of GABA (100 µM) also caused a shortterm (~1 s) efflux of Cl − from the proteoliposomes with a maximal peak of fluorescence change of 13.0 ± 0.5% (n = 4) that was quickly restored and reached a plateau during 60 s. Both the first and repetitive application of the agonist were inhibited by γPAs, showing the involvement of ATPase activity ( Figure 5B,C). To examine the role of ATPase in the GABA-mediated Cl − influx into proteoliposomes, we studied its activity in the absence or presence of γPAs. The reconstituted receptor showed Cl − , HCO 3 − ATPase activity in 4.1 ± 0.3 µmol −1 P i min −1 mg −1 ( Figure 5D). In addition, AMP-PNP (2 mM) and VO 4 3− (20 µM) suppressed the Cl − , HCO 3 − ATPase activity. An application of NEM (300 µM) increased the enzyme activity (6.4 ± 0.4 µmol −1 P i min −1 mg −1 ) and this effect was eliminated by the addition of VO 4 3− (20 µM) ( Figure 5E).

Discussion
Although the involvement of kinases and other molecules in the desensitization and slowdown of deactivation of GABA A R-mediated currents has been shown [9,20], the role of the ATP-hydrolyzing system in these processes is less explored. Currently, three fundamental states have been established that depict the GABA A R channel function at sustained or repeated application of agonist binding: resting, open, and desensitized states [2,16]. Here, by repeated use of the agonist, we also observed three forms of GABA A R-mediated Cl − influx in the neurons that vary in their sensitivity to γPAs ( Figure 6). The open GABA A R state, in contrast to the desensitization state, was activated by γPAs ( Figure 1) which is similar to data from other studies. Specifically, blockers of phosphatases (vanadate or okadaic acid) in the concentration of 100 µM [20] as well as AMP-PNP (1 mM) [21,25] activated the GABA A R-mediated Cl − current via changes of processes of phosphorylation and dephosphorylation [46]. However, we found that in the presence of physiological concentrations of HCO 3 − , blockers as well as AMP-PNP inhibited the open state of the GABA A Rs that were shown to take part in the ATP-consuming system in the processes of desensitization and resensitization (Figure 2). This assumption was also supported by the data about changes in [ATP] i in the presence of a physiological concentration of HCO 3 − and the alkylating agent NEM (Figure 3). AMP-PNP, which traps the magnesium nucleotide within the ATP-hydrolyzing site [30], effectively eliminated the GABA A R-mediated Cl − /HCO 3 − fluxes. Similar conclusions were drawn from experiments in which the GABA A R conductance was maintained by the addition of Mg 2+ -ATP (2 mM) not revealed in the presence of non-hydrolysable ATP analogs (AMP-PNP or AMP-PCP) [22][23][24][25].
GABAergic neurotransmission has the unique property of ionic plasticity, rooted in the short-term and long-term changes in the neuronal intracellular concentrations of chloride ([Cl − ] i ) and bicarbonate ([HCO 3 − ] i ) ions [47,48]. Although GABA A Rs are mainly permeable to Cl − , they also show a significant permeability to HCO [49,50]. Experimental studies have demonstrated that massive GABAergic stimulation can shift E GABA from hyperpolarizing toward depolarizing and even excitation [42,44]. In contrast to Cl − , much less is known about the role of HCO 3 − , which despite a lower permeability, invariably flows outward after significant GABA A R activation with a consequent dramatic fall in intracellular pH (pH i ) [41,[50][51][52]. Moreover, it was shown that the reapplication of GABA (100 µM) does not reveal the desensitization of bicucullineinhibited and GABA A R-mediated HCO 3 − outflux in in situ experiments [43,50]. Here, we showed that the presence of HCO 3 − not only increased the [Cl − ] but also reduced pH i after HCO 3 − outflow via the GABA A R pore. We also did not find that the desensitization of GABA A R-mediated HCO 3 − outflux at the second application of the agonist (Figure 3). In addition, such recovery of [HCO 3 − ] i after GABA A R-mediated activity occurring at 40 s is consistent with data on the recovery of GABA A R-mediated conductance after desensitization at~30 s [41,51].
Although the involvement of kinases and other molecules in the desensitization and slowdown of deactivation of GABAAR-mediated currents has been shown [9,20], the role of the ATP-hydrolyzing system in these processes is less explored. Currently, three fundamental states have been established that depict the GABAAR channel function at sustained or repeated application of agonist binding: resting, open, and desensitized states [2,16]. Here, by repeated use of the agonist, we also observed three forms of GABAARmediated Cl − influx in the neurons that vary in their sensitivity to γPAs ( Figure 6). The open GABAAR state, in contrast to the desensitization state, was activated by γPAs ( Figure  1) which is similar to data from other studies. Specifically, blockers of phosphatases (vanadate or okadaic acid) in the concentration of 100 μM [20] as well as AMP-PNP (1 mM) [21,25] activated the GABAAR-mediated Cl − current via changes of processes of phosphorylation and dephosphorylation [46]. However, we found that in the presence of physiological concentrations of HCO3 − , blockers as well as AMP-PNP inhibited the open state of the GABAARs that were shown to take part in the ATP-consuming system in the processes of desensitization and resensitization (Figure 2). This assumption was also supported by the data about changes in [ATP]i in the presence of a physiological concentration of HCO3 − and the alkylating agent NEM (Figure 3). AMP-PNP, which traps the magnesium nucleotide within the ATP-hydrolyzing site [30], effectively eliminated the GABAAR-mediated Cl − /HCO3 − fluxes. Similar conclusions were drawn from experiments in which the GABAAR conductance was maintained by the addition of Mg 2+ -ATP (2 mM) not revealed in the presence of non-hydrolysable ATP analogs (AMP-PNP or AMP-PCP) [22][23][24][25].  Unlike the various plasmalemmal ATPases that use the ATP energy for actively transporting ions against an electrochemical gradient [53], GABA A Rs have passive conduits for anions [1]. Our studies also demonstrated that the GABA A Rs have passive permeability for Cl − or HCO 3 − (Figures 1 and 3). The homomeric β3 GABA A R subtype operates in two different modes, which was shown before in the presence of HCO 3 as a GABA-gated ion channel or a primary-active ATP-consuming transporter [27]. Such properties are characteristic of several systems, including intensively studied systems such as ATP-binding cassette importers [54] or exporters [55] inhibited by high concentrations (≥100 µM) of vanadate or okadaic acid [56,57]. However, detailed mechanistic insights into the role of ATP consumption in the function of these systems are lacking and most probably involves in channel gating [58]. P-type ATPases are characterized by the formation of transit high-energy acyl phosphate intermediates, where ATP is bound at the catalytic site as a planar structure in a complex with water and Mg 2+ in a dephosphorylation transition state-like conformation that is inhibited by low concentrations (≤20 µM) of VO 4 3− or metal fluoride (MeF x ) complexes [56,57]. Previously, we showed that the GABA A R-coupled ATPase can form a phosphate intermediate that is dephosphorylated in the presence of anions and inhibited by low concentrations of VO 4 3− (≤20 µM) [28]. Phosphorylation of the β3 subunit was seemingly an intermediate step required for energy transduction and displacement of bound anions before hydrolytic cleavage of P i during the resensitization state (Figures 1 and 2). In addition, the NEM effect on the GABA A R-mediated Cl − or HCO 3 − fluxes was eliminated by vanadate in the presence of HCO 3 , which denoted a close site of localization of the ATP-hydrolyzing center and cysteine residue (C313) in the M3 domain of the β3 subunit ( Figure 6). This line of reasoning confirms data where the chimeric isoform did not show the desensitization of the GABA A R-mediated Cl − flux but only the resensitization.
Although stereotypical synaptic GABA A Rs are composed of two α, two β, and one γ subunit [59], the functional properties of GABA A Rs are ensured by the mandatory inclusion of the β subunits and primary β3 subunit [60]. Furthermore, the β3 subunit possesses a set of functional and pharmacological properties that distinguish it from other β subunits [28,61]. These observations lend a new physiological significance to the β3 subunit in the manifestation of resensitization of GABA A R via the involvement of the ATPase. The changing [HCO 3 − ] i is the major determinant of the appearance of GABA A R resensitization, whereas the changing [Cl − ] i in the bicarbonate-free medium plays a vital role in their desensitization that is similar to data of other studies. In the particular, it was found that the changing [Cl − ] i is the major determinant of GABA A R-mediated current decay in the presence of an agonist [11,62]. Overall, we describe a new investigation of the role of HCO 3 − in the GABA A R function and revealed a molecular mechanism whereby a massive activation of GABA A Rs can strengthen or reduce the inhibition via GABAergic synapses. In this context, given the current structural and kinetic evidence, the physiological role of the ATPase appears to be essential due to its involvement in the long-lived conformational change of receptors, transduced from the external to the internal faces of the plasma membrane upon the binding of an agonist. However, structural studies are required to establish precisely what kind of molecular rearrangements take place in the structure of the β3 subunit during the transition from desensitization to resensitization.

Animals and Housing
Animal experiments were carried out using adult male Wistar rats purchased from the Institute of General Pathology and Pathophysiology vivarium and weighing 130-160 g at the time of arrival unless otherwise stated. Rats were always group-housed (5 per cage) and maintained in a temperature-controlled environment (23 ± 1) on a 12:12 h light-dark cycle and had access to food and water ad libitum. We performed all manipulations on animals in accordance with EU directive 2010/63/EU and according to the principles expressed in the

Synaptoneurosomes (SNs) Preparation
SNs were prepared from whole brains of wild-type from freshly dissected forebrains (mostly cortex) (~200-400 mg wet weight) as previously described [63]. Briefly, rats were quickly decapitated using a guillotine, brains were removed and placed in an ice-cold, balanced salt solution (BSS) containing 135 mM NaCl, 1 mM KCl, 0.8 mM MgCl 2 , 0.5 mM KH 2 PO 4 , 10 mM glucose, 0.1% bovine serum albumin (BSA), 10 mM Hepes-Tris (pH 7.3), and a protease inhibitor (A32955, Thermo Fisher Scientific, USA). The brain was cut into small pieces (2-3 mm) and manually homogenized (6 strokes) with a loosely fitting glass-Teflon homogenizer. The homogenate was passed through a nylon mesh (80 µm), and the filtrate was subsequently passed through a cellulose nitrate filter (8 µm) followed by centrifugation at 1000× g for 15 min. The pellet was washed once in BSS and centrifuged. All the procedures were performed at 4 • C. Sodium chloride (7647-

Cell Cultures and Transfection
For the expression homo-or heteromeric GABA A R ensembles, human embryonic kidney 293FT cells (American Type Culture Collection) were used. The cells were purchased from Invitrogen (USA) as part of the Membrane Pro™ Functional Protein Expression System (A11667), and the cell line identity was not further authenticated. The cells were grown and maintained in an incubator (Sanyo, Osaka, Japan) at 37 • C in a humidified atmosphere with 5% CO 2 , in DMEM media (41965-039, Gibco, Inchinnan, UK) supplemented with of 0.1 mM MEM NEAA (11140035, Gibco, Inchinnan, UK), 4 mM L-glutamine, 1 mM sodium pyruvate, 4.5 g/L D-glucose (15023021), and 10% FBS ((10270-106, Gibco, Germany) until the 20th passage, as suggested by the vendor. HEK 293FT cells were transfected by Lipofectamine TM 2000 or 3000 (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA and Lithuania) transfection reagents according to the manufacturer's instructions. Cells were harvested and analyzed 24 h after transfection. For transfection procedures and virus-like particle (VLP) production, the same growth medium with decreased FBS content up to 4% was used according to the manufacturer's recommendations. Geneticin G418 sulphate (11811031, Invitrogen, Waltham, MA, USA) was present in the growth medium at a concentration of 500 mg/mL constantly except during the transfection. The cells were subcultured at confluence by treatment with 0.05% trypsin and 0.02% EDTA in PBS. For selection purposes and improving the yield of VLPs, the transfection medium was removed after 24 h and a fresh growth medium with 10 µg/mL blasticidin (R21001 Gibco, USA) was added. Transfected cells and VLPs were collected and analyzed 24-48 h after transfection.

Molecular Biology
The genes encoding the full-length rat GABA A R β3 subunit were amplified by PCR from the cDNA library (Evrogen, Moscow, Russia) using gene-specific primers with Kozak sequence at the 5 end of the forward primer based on "GenBank: NM_017065.1" sequences. The PCR products were cloned into the pEF6/V5-His TOPO TA vector (K961020, Invitrogen, Waltham, MA, USA) separately and verified by DNA sequencing. Each vector was amplified using E. coli TOP10 strain in LB medium supplemented with 20 µg/mL ampicillin. Isolation and purification of plasmids were performed with PureYieldTM Plasmid Miniprep System (Promega, Madison, WI, USA) and Plasmid Midiprep 2.0 (Evrogen, Moscow, Russia). The sterilization of plasmids was implemented via 0.22 µm filtration. The concentration of plasmids was evaluated on spectrophotometer NanoDrop 1000 (Thermo Fisher Scientific, Waltham, MA, USA). The quality validation of cloning and growth was performed additionally through enzymatic restriction by XbaI and BamHI in BamHI buffer (Thermo Fisher Scientific, Waltham, MA, USA), and the following electrophoresis in 1% agarose gel.

Transfection
The typical transfection procedure of GABA A R subunit-containing constructs for the subsequent biochemical, spectrofluorometric, and Western blot analyses was as follows. Approximately 5 × 105 HEK293FT cells were suspended in 8 mL DMEM, plated into a 90-mm culture dish, and maintained 24 h approximately until 50% to 90% of the confluence. Then, 5 µg of plasmid DNA (β3 alone) was added combined with Lipofectamine ® 3000 Reagent (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA) in Opti-MEM ® I (1×) + GlutaMAXTM-I medium (51985-026, Gibco, Inchinnan, UK) accordingly the manufacturers' recommendations. For microscopy, the cells were plated in 35 mm dishes and were incubated with a proportional amount of reagents and vectors.

VLP Production
For VLP production, GABA A R subunit-containing constructs were transfected together with Membrane Pro™ Reagent (Invitrogen, Waltham, MA, USA) amenably. Transfected HEK293FT started to bud off VLPs from the cell membrane approximately 24 h after transfection. The harvesting procedure was executed in conformity with manufacturer's recommendations. Briefly, the VLP-containing medium was mixed with Membrane Pro™ Precipitation Mix in the ratio of 5 to 1, where 5 refers to the medium. Then, the mix was incubated at 4 • C overnight. After incubation, VLPs were pelleted by centrifugation at 5500× g for 30 min and resuspended in HEPES buffer for subsequent analysis or stored at −80 • C.

Cl − and HCO 3 − -Transport Assays
Cl − -sensitive fluorescent dye MQAE (N-(Ethoxycarbonyl-methyl)-6-Methoxyquinolinium Bromide) or BCECF, AM (2 ,7 -Bis-(2-Carboxyethyl)-5-(and-6)-Carboxyfluorescein, Acetoxy methyl Ester) were obtained from Thermo Fisher Scientific, USA (E3101) or B1170, respectively. A stock solution was prepared in H 2 O, aliquoted, and stored in the freezer (−20 • C) protected from light. For cases in which fluorescence measurements were conducted, the HEK 293FT cells or SNs with loaded dye were stored in an opaque test tube at RT or 4 • C. Sodium bicarbonate (144-55-8), γ-aminobutyric acid, GABA (56-12-2), and bicuculline (485-49-4) were obtained from Sigma-Aldrich (USA), and adenosine 5 -(β,γ-imido)triphosphate lithium salt hydrate (25612-73-1) were obtained from Merck (USA). GABA A R-mediated Cl − -or HCO 3 − -transport was assessed by the dynamic measurements of the variations in the fluorescence intensity of Cl − -sensitive fluorescent dye MQAE-loaded or BCECF-loaded HEK 293FT cells or SNs using a FluoroMax ® -4 spectrofluorometer (HORIBA Scientific Edison, Piscataway, NJ, USA), respectively. For that, the control HEK 293FT cells and various GABA A R β3 isoforms cells were trypsinized by adding 0.05% trypsin-EDTA solution (25200056, Gibco BRL, USA), washed PBS twice, resuspended in the BSB, and then loaded with MQAE for 1 h at 37 • C. After loading, the suspension was centrifuged at 200× g for 5 min at RT and kept in the aforementioned medium at RT in the opaque test tubes. For analysis, the pellet was resuspended in the BSB. Dye-loaded cells (SNS or HEK 293FT) or proteoliposomes were equilibrated in the incubation medium in the absence or presence of compounds (γPAs, bicuculline, or NEM) for about 10 min at 37 • C before initial fluorescence measurements, and then 150 µL of the suspension was added into quartz microcuvette (non-flow cell) andstirred. The GABA-mediated Cl − or HCO 3 − transport was initiated directly in the cuvette by an in-house solution supply system. The excitation and emission wavelengths were 350 nm and 480 nm for measurement of Cl − -transport or 490 nm and 535 nm for measurement of HCO 3 − -transport, respectively. The ∆F/F of each trial was calculated as (F − F 0 )/F 0 , where F 0 is the baseline fluorescence signal averaged over a 25 s period (this was the control measurement) immediately before the start of the application of GABA and supplement compounds. The value of 100% was obtained as the fluorescence intensity before the application of GABA, in the absence or presence of test compounds. The maximum amplitude of GABA-mediated fluorescence responses (first peak % and second peak %) was calculated as the maximal difference in fluorescence intensity in the absence or presence of an agonist.

[ATP] i Measurement
ATP concentrations were determined by the luciferase method. As the source of luciferase, we used the ATP-Glo luminescence kit from Biotium (Fremont, CA, USA). The ATP standards or samples of SNs were prepared in 100 µL dH 2 0 for the assay. The solution containing luciferase was reconstituted as indicated by the manufacturer and was stored in aliquots at −20 • C for at least 2 weeks. For measuring ATP, the solution was diluted 25-fold with an ATP assay mix dilution buffer. A calibration curve with 2-200 pmol of ATP was prepared by diluting an ATP stock solution (2 mM). ATP standards or samples were added to give a final volume of 500 µL. Tubes were vigorously vortexed for 3-5 s and the content was immediately placed in the 96-well black plate and then in the multilabel counter (Plate ChameleonTM V, Finland). The procedure lasted 3-5 min. D-luciferin was dissolved to a concentration of 0.4 mg/mL. Firefly luciferase was added to the ATP-Glo assay solution at a rate of 1 µL per 100 µL (25 µL luciferase per 2.5 mL). The ATP-Glo detection cocktail was prepared fresh before each use for maximum potency. The integration time of the luminescent signal was 10 s. As well, 100 µL of ATP-Glo detection cocktail was added to the sample. We mixed the mixture rapidly by pipetting. Immediately thereafter, the luciferase activity was measured at room temperature.

ATPase Activation Assay
The Cl − /HCO 3 − -ATPase activity in PMs of neurons or HEK 293FT cells expressing the various constructs were measured as previously described [27]. Briefly, PMs or VLPs were added to 0.5 mL of an incubation medium containing 20 mM HEPES-Tris pH 7.3, 5 mM NaCl/25 mM NaHCO 3 , and 50 mM NaNO 3 (neutral salt) to measure enzyme activity. The enzyme preparations (~10 µg) were preincubated at 37 • C for 15 min with the relevant compounds in an incubation medium containing 20 mM HEPES-Tris pH 7.3 and 5 mM NaCl/25 mM NaHCO 3 . Preparation of the test tube with bicarbonate-NaHCO 3 (1 mM)-was previously dissolved in HEPES (20 mM) and then added in the 20 mM HEPES-Tris buffer (pH 6.7). The reaction was started by addition of Mg 2+ -ATP 2 mM (final concentration) in an experimental medium. The Cl − /HCO 3 − -activated ATPases activity was determined as a difference in formation of inorganic phosphorus (P i ) in the absence and in the presence of 5 mM NaCl/25 mM NaHCO 3 in the incubation medium. Adenosine 5 -triphosphate (ATP) disodium salt hydrate (34369-07-8) and adenylyl-imidodiphosphate (25612-73-1) were obtained from Merck (USA). The concentration of inorganic phosphate (P i ) in the incubation medium was measured by a modified method of Chen et al., (1956) [27] using a Cary 60 UV-vis spectrophotometer (Agilent, Santa Clara, CA, USA) at wavelength of 650 nm. The γ-phosphate analog, orthovanadate (VO 4 3− ) (Sigma-Aldrich, St. Louis, MI, USA), was obtained by boiling the vanadate solution ([pH 10]; 10 min), and freshly boiled stock was diluted to the final concentration (pH 7.3) prior to use.