GABA_A receptors activate fish feeding behaviour via two distinct functional pathways

Over recent decades there has been tremendous progress in deciphering how the nervous system modulates feeding behaviour and appetite. Cellular and molecular studies demonstrated that γ-aminobutyric acid (GABA), a major inhibitory neurotransmitter, plays a critical role in feeding behaviour regulation (Williams and Elmquist, 2012; Sohn et al., 2013). When activated, ionotropic GABA_A receptors (GABARs) commonly induce appetitive increases and upregulate feeding activity (Reynolds and Berridge, 2001; Wu et al., 2009).

Benzodiazepines, which act as allosteric GABAR agonists, have long been accepted as a class of compounds that induces hyperphagia and activates appetite after binding to GABARs. This effect has been established in a number of animal species (Brown et al., 1976; Baile and McLaughlin, 1979; Cooper et al., 1985; Gaskins et al., 2008) and in humans (Haney et al., 1997), where benzodiazepines were shown to control abnormal feeding activity and obesity (Fieldstone et al., 1998).

In contrast to Tetrapoda, the effect of benzodiazepines in lower animal taxa is much less studied. In particular, to the best of our knowledge, there are no data on changes in fish feeding behaviour induced by benzodiazepines, despite this class of compounds having been reported to upregulate overall food consumption when continuously present in water (Belal and Assem, 2011; Brodin et al., 2013). In addition, to the best of our knowledge, there is no or little information on the behavioural repercussions of the interaction between benzodiazepines and other GABAR ligands in fishes.

A number of marine and freshwater fish species were introduced as experimental models in food consumption research, due to their relatively simple and easily quantifiable feeding behaviour (Yamashita et al., 2006; Matsuda, 2009; Facciolo et al., 2010). In particular, Carassius gibelio (Prussian carp) has been established as a popular animal model for research of GABAR-mediated effects on behaviour (Facciolo et al., 2011; Volkoff, 2013), which provides a rationale for our choice of this species as an experimental subject.

A contemporary approach to in vivo receptor modulation and drug design implies combined usage of compounds with different types of molecular action mechanisms: orthosteric (competing for binding site with endogenous ligand) and allosteric (not competing with endogenous ligand). Combinations of different ligand types typically afford unprecedented levels of selectivity and fine tuning of end effects (Christopoulos, 2002; Wenthur et al., 2014). Thus, due to the allosteric action mechanism of benzodiazepines, an important research direction regarding their pharmacological properties is investigation of mutual modulation between benzodiazepine effect and effects caused by GABAR orthosteric ligands.

Taken together, these facts imply the need for behavioural study combined with research of benzodiazepine effects at the cellular and molecular level to clarify benzodiazepine-controlled molecular mechanisms of GABAR modulation and to correlate them with a profile of behavioural response. As this approach aggregates data generated at different levels of organization (cellular, molecular and whole organism), it provides an opportunity for a complex characterization of perspective chemical compound(s) that have not yet been fully explored.

Phenazepam [7-bromo-5-(2-chlorophenyl)-1,3-dihydro-2H-1,4-benzodiazepine-2-one] (PNZ), is a benzodiazepine that was synthesized in the USSR in 1974 (Bogatskii et al., 1975) and was never in clinical use outside of former Soviet territory. There are scarce data available on PNZ pharmacology, pharmacokinetics and molecular mechanisms of action, and even less about its impact on behaviour (Maskell et al., 2012); however, it was proven to exert a significant modulatory effect on GABAergic inhibitory currents (Kopanitsa et al., 2002) and weaken chemical stress reactions in fishes (Potrohov, 2008). Therefore, we selected PNZ for our study as an under-examined allosteric agonist allowing modulation of GABAR-conducted currents and promising a significant impact on the behavioural profile of C. gibelio. We set out to test whether and to what extent fish feeding behaviour is controlled by PNZ action, to relate these data to cellular and molecular effects of benzodiazepine binding site activation, and to clarify, in this context, the functional interaction of modulatory effects induced at orthosteric (GABA-binding) and allosteric (benzodiazepine-binding) sites of GABAR.

All experiments were performed on wild-type C. gibelio (Bloch 1782) (Prussian carp), body length 10–15 cm, of either sex. Fish were maintained, and all experiments were performed at room temperature (22–25°C). Before treatment fish were allowed to acclimate for 2 weeks under a 12 h light:12 h dark photoperiod in flow-through constantly aerated tanks (60 litres). Fish were fed once a day with commercial granulated food. Animal maintenance and experimental procedures were conducted in accordance with the Guide for Use and Care of Laboratory Animals (European Communities Council Directive, 1986) and UK Animals (Scientific Procedures) Act 1986.

After the acclimation period, fish were randomly placed into testing tanks shielded by white paper sheets to minimize external disturbance. Subsequently, the animals were lightly anaesthetized by placing them in iced water (Facciolo et al., 2011) and injected intraperitoneally with GABAR ligands. Injections were administered caudal to the pelvic fin. To verify that the injections themselves did not influence feeding behaviour, we performed a control test where unhandled fish (N=6) and fish injected with 0.1 ml of 0.9% NaCl (N=6) were compared by feeding behaviour over 8 h. On the basis of similar behavioural responses (average number of feeding acts) between groups, saline-treated animals were considered as controls for all behavioural experiments. One further control test was performed to determine when fish could be returned to experimental use. In this test we compared feeding behaviour of a control fish group and fish groups subjected to injections of 0.3 mg kg⁻¹ PNZ, 3 mg kg⁻¹ bicuculline (Bic) or 3 mg kg⁻¹ SR-95531 (SR). No difference in feeding act number was found after 3 days post-injection. We therefore used each fish for injection of a single dose of GABAR ligands 4–6 days after the end of the previous experiment.

Feeding behaviour was monitored and quantified as described previously (Volkoff, 2013). Briefly, a ‘feeding act’ (FA) was defined as an event when the fish consumed food granules (FAC), or engulfed a granule and then ‘spat’ it out (FAS), or pushed a granule with closed mouth (FAP).

For the evaluation of the GABAR effects on the feeding behaviour (number of FAs), fish were injected with GABAR ligands diluted in 0.9% NaCl vehicle and compared with animals that received an i.p. injection of vehicle only (controls). Episodes of convulsive behaviour after injection of GABAR antagonists were defined as those where intermittent rapid swimming in circles and/or sporadic rapid swimming with abrupt changes of direction were observed (Facciolo et al., 2011). Immobility episodes after injection of benzodiazepine were defined as those where stillness was observed for ≥10 min (Ide and Hoffmann, 2002).

To prepare brain slices, fish were anaesthetized with 130 mg l⁻¹ MS-222 (Posner et al., 2013), and then rapidly decapitated. The brain was rapidly removed, placed into agar block and transferred to a Leica VT1000 vibratome slicing chamber. For slice preparation, we used telencephalon as a part of fish brain which is tightly connected to feeding behaviour (Lin et al., 2000), and where GABAR response in C. gibelio was clearly demonstrated (Hossein-Javaheri et al., 2017).

Electrophysiological patch-clamp recordings were performed with a MultiClamp 700B amplifier (Molecular Devices) under Clampex 10x software, using 4–6 MΩ borosilicate glass electrodes at room temperature (22–25°C). Perfusion media contained (mmol l⁻¹): 108 NaCl, 2.5 KCl, 2 MgCl₂, 2.5 CaCl₂, 24 NaHCO₃, 1 NaH₂PO₄ and 12 glucose; gassed with 95% O₂ and 5% CO₂, pH adjusted to 7.4 with NaOH. Slicing media differed from perfusion media by concentrations of CaCl₂ (0.5 mmol l⁻¹) and glucose (5 mmol l⁻¹). Intracellular electrode solution contained (mmol l⁻¹): 115 CsCl, 25 Hepes, 3 Mg-ATP, 0.5 Na-GTP, 0.5 EGTA and 10 tetraethylammonium chloride; pH 7.2. To isolate GABAR response we added to perfusion solution (μmol l⁻¹): 50 APV, 20 NBQX, 250 MCPG, 1 CGP-55845 and 1 strychnine. GABAR ligands (SR-95531, bicuculline, phenazepam, flumazenil and picrotoxin) were added to the perfusion solution during the course of the experiment. SR-95531, bicuculline, picrotoxin, strychnine, MCPG, NBQX, CGP-55845, APV and flumazenil were purchased from Tocris Bioscience; phenazepam was synthesized and kindly provided by the Bogatskii Physico-Chemical Institute (Odessa, Ukraine); all other chemicals were purchased from Sigma-Aldrich.

Charge transfer per second in single-channel recordings from outside-out patches was calculated as I/(UΔt), where I is the amplitude of a single-channel response (pA), U is the holding membrane potential (mV), and Δt is the time spent in an open state (ms). Time intervals when GABARs were open were selected using a detection threshold of 1.5 pA more negative than mean baseline and a minimum opening time of 0.2 ms. Values for the open time of single channels were obtained with threshold-detection algorithm of Clampfit 10x software.

Non-linear fitting and statistical calculations were performed with Mathematica-10 and Mathematica-11 software packages. Data are presented as means±s.e.m. and evaluated for statistical significance with Student's paired or unpaired t-test, as indicated. Dose–response effects on behaviour elements were assessed with one-way ANOVA; in time–response behavioural experiments, two-way ANOVA followed by a Bonferroni post hoc test was used for comparison between groups.

To determine the most effective doses for GABAR ligands, we performed a ‘dose–response’ study (Fig. 1). In dose–response experiments with a 2 h observation interval starting 15 min after drug injection, we aimed to find a GABAR ligand dose that produced the strongest effect without immobility episodes (PNZ) or episodes of convulsive behaviour (GABAR antagonists). The effect of GABAR ligands was tested at overall number of FAs and separately for FAC, FAS and FAP number; see Table 1 for one-way ANOVA results.

PNZ demonstrated a significant gradual increase of number of FAs with increasing dose (Fig. 1A), but without significant effect on any of the separate FA components (FAC, FAS, FAP). A dose of 0.2 mg kg⁻¹ was the highest at which fish did not demonstrate immobility episodes; higher doses did not generate a number of FAs significantly higher than that for 0.2 mg kg⁻¹. Competitive GABA antagonist SR-95531 (SR) and inverse agonist with partial allosteric effect bicuculline (Bic) caused a decrease in FA number with increasing dose; the highest effect without convulsion episodes was produced by 2 mg kg⁻¹ SR and 1.5 mg kg⁻¹ Bic (Fig. 1B,C). In both cases FAC and FAS, calculated alone, were found to be significantly influenced by incremental doses of GABAR ligands (Table 1). Flumazenil (FMZ; allosteric antagonist acting at the benzodiazepine binding site), injected in doses up to 5 mg kg⁻¹, did not generate a significant effect on overall FA number as well on FAC, FAS and FAP components calculated separately (Fig. 1D, Table 1). Therefore, for further in vivo experiments, we used doses of 1.5 mg kg⁻¹ for Bic, 2 mg kg⁻¹ for SR and 0.2 mg kg⁻¹ for PNZ. FMZ, being injected alone, did not display significant effects at any dose; thus taking into account data from the literature [dosage of up to 40 mg kg⁻¹ in in vivo experiments (Petry and Heyman, 1997)], we have chosen a highest tested dose of FMZ (5 mg kg⁻¹) for further behavioural experiments. As the overall number of FAs was the only parameter that demonstrated significant dependence from the three GABAR ligands (Table 1), we used it as an assessment criterion in further tests.

To find the optimal time course for combined injections of GABAR ligands, we performed a ‘time–response’ behavioural experiment with the optimal ligand doses determined earlier (Fig. 2A). We found that PNZ developed a maximum effect 1.5 h after injection, whereas SR and Bic did so within 1 h; again, FMZ did not display significant difference from control during all the periods of observation.

Next, we tested combined effects of GABAR ligands on C. gibelio feeding behaviour. To do this, we first injected PNZ and, 0.5 h later (to overlap peak responses), 1.5 mg kg⁻¹ of Bic, plus different doses of SR (Fig. 2B). Unexpectedly, a low dose of SR reversed Bic antagonism to PNZ: F_1.90=4.22, P=0.042, two-way ANOVA for 0.3 mg kg⁻¹ SR; whereas a higher dose (0.5 mg kg⁻¹) potentiated the Bic effect. To find out if this paradoxical observation was due to modulation caused by the allosteric site after binding of PNZ, we repeated the experiment without PNZ pre-injection (Fig. 2C). In this case, incremental doses of SR monotonically enhanced the Bic effect in a dose-dependent manner. SR at 0.3 mg kg⁻¹ did not cause a significantly different effect from that evoked by Bic only: F_1.96=3.71, P=0.057, two-way ANOVA. However, 0.3 mg kg⁻¹ of SR evoked convulsive episodes in four out of six experimental animals; thus we did not continue the experiment with higher SR doses.

Dependence of Bic+SR effects from the presence of PNZ confirmed involvement of the benzodiazepine binding site into reversal of the Bic effect caused by SR. To further test this, we performed an experiment where 1.5 mg kg⁻¹ of Bic and incremental doses of SR were injected 0.5 h after combined injection of PNZ and FMZ (Fig. 2D). Here Bic plus 0.3 mg kg⁻¹ SR generated a significantly higher response than Bic alone (F_1.90=6.86, P=0.01, two-way ANOVA); convulsive episodes were observed in three out of six animals. Again, in this experiment SR enhanced the Bic effect in a dose-dependent manner, thus resembling data from the experiment where SR and Bic were used without PNZ.

Equipped with this knowledge, we formulated a working hypothesis explaining the effects of SR, Bic and PNZ as follows. SR and Bic, applied together, compete for the same binding site with endogenous GABA; their joint effect is thus proportional to the number of occupied GABA-binding sites at GABARs and, therefore, increases continuously when SR concentration increments. When applied after PNZ, Bic evokes both a competitive effect (displacing GABA at a conjoint binding site) and an allosteric effect (counteracting conformational changes induced by PNZ). In contrast, SR in small doses, after dislodging of Bic from the binding site, delivers only competitive impact without counteraction to the allosteric effect of PNZ, thus weakening joint antagonist effect on GABARs; however, incrementally increasing doses of SR occupy additional GABA-binding sites and, thus, start to enhance the effect of Bic.

Therefore, we proceeded to carry out electrophysiological experiments to test this working hypothesis. To do this, we aimed to dissect the effects of PNZ, Bic and SR and their mutual modulation at the whole-cell and single-receptor level.

In the first part of the electrophysiological section of our study, we performed recordings of spontaneous synaptic GABAergic responses in whole-cell mode. We presumed that the decay profile of spontaneous inhibitory post-synaptic currents (sIPSCs) should consist of two exponents, with one exponent generated by orthosteric (GABA-binding) site effects, and the other by effects of the allosteric (benzodiazepine-binding) site. Thus, for quantification of effects generated at orthosteric and allosteric sites, we used the ratio of fitting coefficients (RFC) of two IPSC decay exponents obtained after fitting of averaged sIPSCs with bi-exponential function (see Fig. 3D legend for more details of analysis paradigm). Picrotoxin (PTX, 50 µmol l⁻¹), applied at the end of each experiment, fully suppressed all synaptic events, thus confirming their exclusive GABAergic origin.

Application of PNZ (0.2 µmol l⁻¹) increased sIPSC amplitudes and significantly slowed down their decay kinetics (Fig. 3A,C); RFC was 3.62±0.64 for PNZ versus 6.03±0.68 in control, P=0.016, N=6, Student's paired t-test. SR (50 nmol l⁻¹), added together with PNZ, did not influence response decay kinetics biased by PNZ: RFC was 2.87±0.47 for PNZ+SR versus 3.62±0.64 for PNZ alone, P=0.235, N=6, Student's paired t-test. In contrast, Bic (50 nmol l⁻¹), when applied with PNZ, restored sIPSC decay kinetics to almost control values: 5.56±0.42 for PNZ+Bic versus 3.62±0.64 for PNZ alone, P=0.029, N=6, Student's paired t-test (Fig. 3E). RFC analysis (Fig. 3D) demonstrated that PNZ application magnified the slow decay exponent, which then remained unaltered by SR, but was suppressed by Bic.

This confirmed the different nature of the antagonism to PNZ effect generated by SR and Bic. However, was it due to modulation of the benzodiazepine binding site by one of these antagonists? To clarify this issue, we repeated the experiment on sIPSCs, but with FMZ applied before Bic (Fig. 4). FMZ (2 µmol l⁻¹) added after PNZ reverted the bias of RFC caused by PNZ: 5.73±0.69 for PNZ+FMZ versus 3.51±0.71 for PNZ alone, P=0.0009, N=6, Student's paired t-test. Bic, added to perfusion solution after FMZ, did not cause a significant effect: RFC was 5.05±0.77 for PNZ+FMZ+Bic versus 5.73±0.69 for PNZ+FMZ, P=0.55, N=6, Student's paired t-test (Fig. 4C).

We thus concluded that the effects of Bic on decay kinetics were due to its impact on the benzodiazepine binding site when the site is activated by PNZ; FMZ, displacing PNZ, prevents the effect of Bic.

Next, to test whether Bic generates an allosteric effect when the benzodiazepine site is inactive, we performed an experiment on sIPSCs without benzodiazepine site ligands (Fig. 5). In this case, both SR and Bic did not show a significant impact on s-IPSC kinetics: RFC was 5.29±0.83 for control, 6.21±0.67 for SR and 5.9±0.78 for Bic; P>0.05 for all comparisons, N=6, Student's paired t-test.

Experiments on sIPSC kinetics confirmed our working hypothesis about the interaction of GABAR antagonists (SR and Bic) with an allosteric benzodiazepine binding site. However, in all these tests, the orthosteric site of the receptor was (possibly) activated by endogenous GABA. Thus, to further dissect mechanisms of action of benzodiazepines and GABAR antagonists, and to clarify their input through the GABAR orthosteric site, we proceeded to single-channel recordings from outside-out patches, where GABA presence and/or concentration can be tightly controlled. Here we used an overall charge transfer per second as a quantitative characteristic of output generated by pharmacological manipulations.

We found that GABA (1 µmol l⁻¹), when added to perfusion solution, induced a clear response that was suppressed by both SR (20 µmol l⁻¹) and Bic (20 µmol l⁻¹) to values similar to those obtained for spontaneous openings (Fig. 6A,C). Charge transfer normalized to that generated by GABA only was 0.11±0.03 for control (spontaneous openings), 0.16±0.03 for GABA+SR and 0.09±0.01 for GABA+Bic; P>0.05 for all comparisons, N=6, Student's paired t-test. However, when added to perfusion solution after application of PNZ (without GABA), Bic, but not SR, exerted significant downregulation on charge transfer (Fig. 6B,D): normalized charge transfer for PNZ+SR was 0.98±0.14 versus 0.6±0.09 for PNZ+Bic, P=0.041, N=6, Student's paired t-test.

Finally, we tried to reproduce the pharmacological steps from in vivo experiments with PNZ and FMZ on outside-out patches (Fig. 2B,D). We thus first applied GABA (1 µmol l⁻¹)+PNZ (0.2 µmol l⁻¹) and antagonized their combined effect by Bic (2 µmol l⁻¹) and Bic (2 µmol l⁻¹)+SR (2 µmol l⁻¹) (Fig. 7A,C). In accordance with the in vivo behavioural experiment, SR partially reversed the effect of Bic on charge transfer: normalized charge transfer was 0.28±0.08 for GABA+PNZ+Bic versus 0.52±0.07 for GABA+PNZ+Bic+SR, P=0.002, N=6, Student's paired t-test. Secondly, we applied GABA (1 µmol l⁻¹)+PNZ (0.2 µmol l⁻¹)+FMZ (2 µmol l⁻¹) and antagonized them with Bic (2 µmol l⁻¹) and Bic+SR (2 µmol l⁻¹) (Fig. 7B,D). In contrast to the experiment without FMZ, application of Bic+SR did not cause a significantly different effect from Bic alone: normalized charge transfer was 0.35±0.07 for GABA+PNZ+FMZ+Bic versus 0.31±0.09 for GABA+PNZ+FMZ+Bic+SR, P=0.68, N=6, Student's paired t-test.

We performed a combined behavioural and in vitro electrophysiological study of GABAR effects generated by two distinct mechanisms: orthosteric and allosteric. Correlation of the results obtained in these two types of experiments clearly demonstrates control of GABARs over C. gibelio feeding behaviour delivered through both mechanisms in (partially) competitive mode.

Benzodiazepines, such as PNZ, after binding to GABAR initiate two types of potentiation: orthosteric (GABA-dependent) due to enhancement of GABAR affinity to GABA (Lavoie and Twyman, 1996) and allosteric (GABA-independent) through increase of GABAR opening probability (Birnir et al., 2000). A key point of our research, which all further conclusions depend on, is the nature of SR and Bic action against PNZ orthosteric and allosteric effects. Bic and SR were shown to compete for the same binding site with GABA (Wagner and Czajkowski, 2001; Johnston, 2013). With different experimental approaches, both orthosteric and allosteric action mechanisms were demonstrated for these two compounds (Heaulme et al., 1986; Krishek et al., 1996; Thompson et al., 1999; Ueno et al., 1997; Bai et al., 2001). The inverse agonist impact of Bic was found to suppress GABA-independent GABAR openings (McCartney et al., 2007; Wlodarczyk et al., 2013) and antagonize effects of GABAR allosteric agonists (Ueno et al., 1997; Mathers et al., 2007). In our experiments, Bic, but not SR, exerted significant antagonism to allosteric PNZ effects when drugs were applied to outside-out patches with no GABA in perfusion solution (Fig. 6B,D). In contrast, they both antagonized the effect of GABA (Fig. 6A,C), which implies competitive action at orthosteric GABAR binding sites. We thus concluded that, in our research model, SR displays competitive (orthosteric) action only, whereas Bic generates both orthosteric and allosteric effects.

The three different types of feeding behaviour investigated in our study undergo different modifications as a result of GABAR activation/deactivation (Fig. 1). The sum of all three types of feeding events (FA) was found to be most correlated with GABAR ligand action, whereas individual modes of behaviour (FS, FC, FP) may not be influenced significantly by application of a particular ligand. This probably reflects the fact that the behaviour types under study are closely interrelated, and, as a consequence, easily undergo mutual transition (Helfman, 1986) within general behavioural shift due to GABAR-triggered impact.

An experiment where SR, applied alone, induced a clear decrease in the number of FAs competing with endogenous GABA (Fig. 1B), confirmed the input of an orthosteric functional pathway into feeding behaviour. When applied together, SR weakened the antagonistic impact of Bic on the effect of PNZ in both behavioural tests (Fig. 2B) and single-channel recordings (Fig. 7A,C), whereas FMZ, acting at the allosteric site, precluded such an effect of SR (Figs  2D and 7D). The most straightforward explanation is that Bic antagonizes both orthosteric and allosteric effects of PNZ, whereas SR, after displacement of Bic, blocks the orthosteric effect of PNZ, but leaves its allosteric effect untouched. Hence these two lines of evidence confirm our working hypothesis about the molecular mechanisms of interaction between SR, Bic and PNZ. Paradoxical antagonism of SR to Bic effect, displayed in vivo and in vitro, unequivocally points to allosteric GABAR potentiation as one more functional pathway for modulation of fish feeding behaviour.

In experiments on sIPSCs, PNZ application slowed down overall decay kinetics due to magnification of exponent with a higher τ value (slow decay component, Fig. 3), which is in line with effects of benzodiazepines observed earlier (Poncer et al., 1996). Bic, added after PNZ, shifted RFC due to suppression of the slow decay component. In turn, FMZ, which competes with PNZ for the allosteric binding site, also suppressed the slow decay component (Fig. 4), suggesting that this component is generated as a result of allosteric site activation. As SR was found to deliver its effect exclusively via the orthosteric site of action, one may expect SR, when applied after PNZ, to generate a bias of RFC opposite to that generated by Bic: suppression of the fast decay component with the slow component left intact. However, we did not observe any significant RFC shift generated by SR (Figs 3 and 5). The plausible explanation for this phenomenon is that allosteric activation of spontaneous openings induced by PNZ (see Fig. 6B) is a stable background input, which does not affect IPSC kinetics. In contrast, increased affinity to GABA after PNZ binding affects responses generated by GABA efflux only. Thus if both fast and slow (the latter magnified due to increased affinity to GABA) decay components are generated at orthosteric sites, SR, displacing GABA, should exert proportional impact on both decay components without changes in RFC, which corresponds to our observations.

Depending on cell type and particular compound, benzodiazepines may affect phasic, tonic, or both types of GABAR-induced inhibition (Bai et al., 2001; Nusser and Mody, 2002; Glykys and Mody, 2006; Yamada et al., 2007). Tonic GABAR conductance can be produced by two functional types of receptors: GABA-dependent, i.e. activated by low concentrations of ambient GABA (Farrant and Nusser, 2005), and GABA-independent, i.e. spontaneously opening (McCartney et al., 2007; Wlodarczyk et al., 2013). Thus, only the ability to modulate tonic conductance through spontaneously opening GABARs may prove the GABA-independent, irreducible to increasing affinity of GABAR to GABA, action of given benzodiazepine. In our experiments, we found that PNZ potentiates phasic and tonic responses of GABARs; tonic conductance was potentiated without the presence of GABA (Fig. 6), which resonates with earlier observations on classical benzodiazepine compounds such as diazepam (Birnir et al., 2000). Hence our data prove that PNZ delivers its impact through two types of GABAR potentiation: GABA-dependent and GABA-independent.

Another observation, which appears to some extent counterintuitive, is a difference between the behavioural effect of increasing doses of SR when PNZ was not injected (no significant difference from the effect of Bic only; Fig. 2C) and when PNZ binding was blocked by FMZ (significant difference from Bic only; Fig. 2D). The possible explanation is that in the native brain, the benzodiazepine binding site is neither fully activated nor suppressed, and thus generates a certain amount of spontaneous receptor openings (as registered in outside-out patches; Figs 6 and 7). SR, being added in addition to Bic, enhances the antagonistic effect of Bic when competing for binding site with endogenous GABA, but removes Bic antagonism to partially activated benzodiazepine sites when competing with Bic. These opposite actions prevent development of a statistically significant end effect of certain SR doses. However, when the benzodiazepine site is blocked by FMZ, competition between SR and Bic does not generate any indirect effect connected to riddance of Bic allosteric action, thus allowing unhindered development of antagonism due to GABA displacement by SR. If this is the case, this also implies regulation of feeding behaviour by two types of GABAR signalling: phasic (synaptic IPSCs generated by GABA efflux) and tonic (spontaneous GABAR openings all over the cell surface).

Our study thus has several outcomes. As a fundamental result, we proved the existence of two distinct molecular mechanisms regulating feeding behaviour through GABARs, which act in a partially competitive manner. Due to the relatively conservative nature of GABAR structure in vertebrates (Martyniuk et al., 2007), these two mechanisms with the given interaction should be expected in higher taxons as well as in fishes. If true, this makes C. gibelio a convenient and universal experimental model for further studies of GABAR effects on feeding behaviour. In addition, we filled a data gap in behavioural effects of PNZ which is a prospective, albeit under-researched, compound for treatment of GABAR-connected disorders.

After rapid growth during recent decades, aquaculture now supplies more than half of the total fish for human consumption; in parallel, aquaculture's share of the world's fishmeal consumption has more than doubled (Naylor et al., 2009). Further progress in fish aquaculture is thus tightly related to intensification and optimization of feeding of the cultivated fishes; this is the subject of intense studies, in particular on species of the Carassius genus (Xue and Cui, 2001; Wang et al., 2015). Therefore, the applied outcome of our study highlights benzodiazepines as prospective compounds for use in aquaculture as stimulants of fish feed intake and weight gain.

The authors thank Professor Alexej Verkhratsky (University of Manchester) for valuable suggestions on the preparation of this paper.