High-Dose Benzodiazepines Positively Modulate GABAA Receptors via a Flumazenil-Insensitive Mechanism

Benzodiazepines (BZDs) produce versatile pharmacological actions through positive modulation of GABAA receptors (GABAARs). A previous study has demonstrated that high concentrations of diazepam potentiate GABA currents on the α1β2γ2 and α1β2 GABAARs in a flumazenil-insensitive manner. In this study, the high-concentration effects of BZDs and their sensitivity to flumazenil were determined on synaptic (α1β2γ2, α2β2γ2, α5β2γ2) and extra-synaptic (α4β2δ) GABAARs using the voltage-clamp electrophysiology technique. The in vivo evaluation of flumazenil-insensitive BZD effects was conducted in mice via the loss of righting reflex (LORR) test. Diazepam induced biphasic potentiation on the α1β2γ2, α2β2γ2 and α5β2γ2 GABAARs, but did not affect the α4β2δ receptor. In contrast to the nanomolar component of potentiation, the second potentiation elicited by micromolar diazepam was insensitive to flumazenil. Midazolam, clonazepam, and lorazepam at 200 µM exhibited similar flumazenil-insensitive effects on the α1β2γ2, α2β2γ2 and α5β2γ2 receptors, whereas the potentiation induced by 200 µM zolpidem or triazolam was abolished by flumazenil. Both the GABAAR antagonist pentylenetetrazol and Fa173, a proposed transmembrane site antagonist, abolished the potentiation induced by 200 µM diazepam. Consistent with the in vitro results, flumazenil antagonized the zolpidem-induced LORR, but not that induced by diazepam or midazolam. Pentylenetetrazol and Fa173 antagonized the diazepam-induced LORR. These findings support the existence of non-classical BZD binding sites on certain GABAAR subtypes and indicate that the flumazenil-insensitive effects depend on the chemical structures of BZD ligands.


Introduction
GABA A receptors (GABA A Rs), the major inhibitory neurotransmitter receptors in the nervous system, are complex receptors that are critically involved in numerous physiological and pathological processes. They are heterogeneous pentamers assembled from at least 19 subunits (α 1-6 , β 1-3 , γ 1-3 , δ, ε, π, ρ 1-3 , and θ), and there are potentially dozens of active GABA A R subtypes with distinct distributional and functional characteristics [1][2][3][4]. Ternary receptors composed of two α, two β, and one γ (or δ) subunits are considered to constitute the majority of GABA A Rs. On the other hand, the complexity of GABA A Rs lies in various allosteric modulatory sites [5,6]. Although knowledge concerning the GABA A R structure and its interaction with various ligands is increasing rapidly [5][6][7][8][9][10][11], a complete elucidation of the modulatory mechanisms of GABA A R is still far from being achieved.
As one of the most important GABA A R allosteric modulators, benzodiazepines (BZDs) have wide and versatile clinical applications. At low or high dosages, BZDs, such as diazepam, produce anti-anxiety, anti-seizure, sedation, and anesthetic effects [12]. Conclusive evidence has shown that the anti-anxiety and sedative effects of BZDs are associated with the high-affinity (classical) sites located at the α+/γ− interface of synaptic GABA A Rs [13,14]. However, whether BZD-induced anesthesia is mediated by the classical binding sites remains to be clarified.
In addition to the classical binding sites, several new binding sites of BZDs on GABA A R are proposed [15][16][17][18]. In particular, a transmembrane BZD binding site that is activated by micromolar concentration of diazepam was suggested on recombinant α 1 β 2 γ 2 and α 1 β 2 GABA A Rs. However, these in vitro experiments were conducted using limited BZDs and α 1 -containing GABA A Rs. Previously, we demonstrated the flumazenil-insensitive BZD effects in a series of αβ binary GABA A Rs [19] and in zebrafish larvae [20]. By using the electrophysiological technique, the present study revealed flumazenil-insensitive diazepam modulation on a series of ternary GABA A Rs, which was abolished by both the GABA A R antagonist pentylenetetrazol and Fa173, a proposed GABA A R antagonist targeting the transmembrane site [21]. The effects of structurally differentiated BZDs were further compared to verify the objectivity and universal significance of the non-classical site. Both pentylenetetrazol and Fa173, but not flumazenil, antagonized BZD-induced anesthesia, which is considered to be related to high-dose BZD effects. The present study provides novel evidence supporting the existence of a flumazenil-insensitive mechanism in BZD modulation of GABA A Rs.

Pentylenetetrazol and Fa173 Abolish the Flumazenil-Insensitive Diazepam Effects
To exclude the possibility that the high-concentration BZD effects were a type of nonspecific effect, the effects of 200 µM diazepam were observed in the presence of either the GABA A R antagonist pentylenetetrazol or Fa173, a proposed transmembrane site antagonist [21]. Similarly, on the α 1 β 2 γ 2 , α 2 β 2 γ 2 and α 5 β 2 γ 2 receptors, diazepam produced two-graded potentiation on the GABA-elicited current at 10 and 200 µM. 100 µM PTZ or Fa173 antagonized not only the low-concentration but also the high-concentration effects of diazepam ( Figure 3). These results indicated that the flumazenil-insensitive effects are specifically mediated by GABA A Rs, and possibly via the transmembrane binding sites.

Anesthesia Induced by Diazepam and Midazolam, but Not Zolpidem, Is Resistant to Flumazenil
The high-dose effects of BZD ligands were evaluated in vivo by using LORR as an index of anesthesia. Diazepam, midazolam and zolpidem were chosen to compare the potential flumazenil-insensitive effects of different ligands. All three ligands dose-dependently induced LORR in mice ( Figure 4A,D,G), and an increase in BZD dose led to a decrease in latency to and an increase in duration of LORR. Complete (100%) LORR was caused by diazepam, midazolam and zolpidem at doses of 50, 100 and 50 mg kg −1 , respectively. Flumazenil treatment failed to antagonize LORR induced by diazepam or midazolam. Moreover, flumazenil even prolonged the duration of LORR at some doses ( Figure 4C,F). In contrast, zolpidem-induced anesthesia was significantly antagonized by flumazenil. Flumazenil dose dependently increased the latency (F (5, 53) = 9.82, p < 0.01) and reduced the duration (F (5, 53) = 11.06, p < 0.01) of LORR induced by zolpidem. In addition, the percentage of zolpidem-induced LORR was reduced to 50% by flumazenil at a dose of 1 mg kg −1 ( Figure 4I).

Diazepam-Induced Anesthesia Is Antagonized by Pentylenetetrazol and Fa173
Pentylenetetrazol and Fa173 were further used to determine the specific target for the high-dose BZD effects in living animals. Compared with the control, pentylenetetrazol treatment significantly increased the latency to (t = 2.88, p < 0.01, Figure 5A) and reduced the duration (t = 3.3, p < 0.01) of LORR induced by diazepam, indicating that the highdose BZD effects were mediated via GABA A Rs. Similarly, Fa173 treatment antagonized diazepam-induced LORR, resulting in significant decreases in LORR percentage (p < 0.05, Figure 5B) and duration (t = 2.571, p < 0.05), and an increase in LORR latency (t = 4.247, p < 0.001). This was consistent with the in vitro results of Fa173 and suggested that high-dose diazepam activates the transmembrane binding sites of GABA A Rs.

Discussion
The present study demonstrated that high doses of BZDs produced profound and flumazenil-insensitive potentiation of GABA-elicited currents on certain GABA A R subtypes. Consistent with the electrophysiological observations, the anesthesia induced by diazepam, the BZD phenotype, was resistant to flumazenil, but antagonized by pentylenetetrazol or Fa173. The findings of this study support the existence of a non-classical mechanism in GABA A R modulation that may contribute to BZD-induced anesthesia.
Direct experimental evidence is still lacking concerning BZD concentration around synaptic and extra-synaptic GABA A Rs after systemic administration, although it is important to determine the clinical significance of high-dose BZD effects. Based on the measured plasmic levels and their high lipid solubility, the concentrations of BZDs in the CNS were estimated to reach double-digit micromolar levels [23,24]. It is reasonable to speculate even higher levels of BZDs in specific brain regions and neural circuits under some circumstances of iatrogenic or self-inflicted overdose. Therefore, the local BZD concentrations were very likely to be high enough for flumazenil-insensitive potentiation under some practical conditions, especially in terms of BZD-induced anesthesia and intoxication.
In contrast to the extensive research on the high-affinity modulatory effects of BZD at nanomolar concentrations, very limited studies have addressed the effects and mechanisms of micromolar BZD. Moreover, there are controversial observations regarding the interaction of low-and high-affinity BZD effects. High-concentration flurazepam inhibited the GABA A R modulation mediated by the classical BZD binding site [15,25], while diazepam at concentrations above 20 µM further potentiated the α 1 β 2 γ 2 receptor [15]. Our results are in good agreement with the latter, and extend the finding on diazepam and the α 1 β 2 γ 2 receptor to a series of ligands and receptors. The flumazenil-insensitive high-dose effects of classical BZDs on synaptic GABA A Rs support the existence of non-classical binding sites, precluding the possibility that the micromolar potentiation of diazepam is a non-specific effect. The fact that Fa173 abolishes the high-dose BZD effects further verifies the nonclassical binding sites, possibly located at the transmembrane domain (TMD) of GABA A Rs. However, other non-classical mechanisms may contribute to high-dose BZD effects [26].
The high-dose BZD effects may mean a broader and deeper depression of the central nervous system, which is considered to be related to general anesthesia [27]. In good agreement with previous studies [20], the full occupation of the non-classical binding sites by high concentrations of BZDs resulted in 2-3-fold increases in the maximum effects relative to modulation via the classical binding sites. Furthermore, the high-dose BZD effects are assumed to affect more GABA A R subtypes, as the construction of the nonclassical binding sites does not require the γ subunit [19]. It is also interesting to see that the non-classical binding sites possessed similar dependence on the α subunit compared to the classical binding sites; i.e., α 1 -, α 2 -or α 5 -, but not α 4 -containing receptors were sensitive to BZD modulation. On the other hand, high-dose modulation was observed in classical BZDs but not non-BZD structures (such as zolpidem), although both categories of ligands bind to the classical binding sites of GABA A Rs. Similarly, a recent study has suggested that the structural features of BZD ligands govern their abilities to bind to the etomidate binding site of the GABA A Rs [28].
Flumazenil, which competes with BZDs to bind to the classical binding site, is the only specific therapeutic for BZD intoxication in clinic. Extensive evidence supports the effective antagonism of flumazenil on sedation, anti-anxiety, and anti-convulsion activity induced by relatively low doses of diazepam [29][30][31], while its effects on high-dose BZDs are rarely studied. The present study investigated the effectiveness of flumazenil against BZD-induced anesthesia, using the LORR as an in vivo model representing high-dose BZD effects. Flumazenil failed to effectively antagonize LORR induced by diazepam and midazolam. In particular, LORR duration was not reduced, but even prolonged by flumazenil under some doses. This result, which was consistent with the in vitro results, suggests that there is a flumazenil-insensitive mechanism in BZD-induced modulation of GABA A Rs, and that flumazenil may be inefficient in antagonizing some BZD effects, such as anesthesia, when used against high doses of certain BZDs.
The binding modes and mechanistic effects of BZD ligands were increasingly resolved, as the high-resolution structures of GABA A Rs in complex with BZDs were presented. Strong densities were observed in the TMD of the α 1 β 3 γ 2 receptor, which is also considered the binding site for general anesthetics [9]. Recently, diazepam was demonstrated to share a binding site with anesthetics in the TMD of the α 1 β 2 γ 2 receptor [11]. The binding of diazepam to this site may contribute to stabilization of the TMD and positively modulate the receptor in a similar way to anesthetics. These structural studies of GABA A Rs provided an excellent explanation for the flumazenil-insensitive BZD effects and the non-classical BZD modulatory mechanism.
In conclusion, the present study provided detailed evidence supporting the existence of flumazenil-insensitive BZD effects in a series of GABA A Rs and their potential association with BZD-induced anesthesia. These findings enhance the understanding of GABA A R modulation and BZD pharmacology, and suggest that some classical BZDs produce profound inhibition of brain function by binding to non-classical sites. Furthermore, elaborate analysis of the interaction of ligands with non-classical sites may prompt the development of novel drugs in modulating wakefulness and sleep [21,32].

Two-Electrode Voltage Clamp Electrophysiology
Whole-cell currents were measured using the two-electrode voltage clamp technique. Microelectrodes were filled with 3 M KCl and those with resistance between 1.0-2.5 MΩ were used. Recordings were performed under constant perfusion at room temperature. Currents were amplified with an OC-725C (Warner Instruments, Hamden, USA) and digitized with a Digidata 1440 (Molecular Devices, San Jose, CA, USA) at 100 Hz. In all cases, currents in response to the application of drugs were recorded using Clampex 10.3 software (Axon Instruments, San Jose, CA, USA) and data were sampled at 2 kHz and filtered at 0.5 kHz. A gap-free protocol was applied with the holding membrane potential at −70 mV. Each drug application was followed by a washout in bath solution (approximately 5 min).

Mouse Behavioral Test
Adult male Kunming mice ages 3-4 weeks (19-21 g) were obtained from Beijing Animal Center (Beijing, China). Animals were housed 10 per cage with free access to food and water on a 12 h light-dark cycle. The mice were handled for 2-3 days to adapt to experimental conditions. All experimental procedures were approved by the local ethical committee and the Institutional Review Committee on Animal Care and Use (IACUC of AMMS-06-2017-003).
The anesthetic effects of BZDs were measured using a mouse LORR model. The mice were tested individually in a clear plastic cage (40 × 20 × 20 cm, l × w × h). After BZD injection, the mice were gently placed in the supine position and the righting reflex was assessed every 1 min until the occurrence of LORR, which manifested as the failure to right themselves within 60 s. The anesthetized mice were left undisturbed until they spontaneously turned over themselves to prone position. Absolute recovery was defined as the mice being able to right themselves twice or more within 60 s. The time between BZD injection and occurrence of LORR was recorded as latency to LORR, and LORR duration was measured as the time from the occurrence of LORR to recovery. The LORR was considered absent if the mice were able to right themselves during the 120 min observation period after BZD injection.

Statistical Analysis
Electrophysiological data were analyzed with Clampfit 10.3 (Axon Instruments, San Jose, USA), Origin 8.0 (OriginLab Corporation, Northampton, MA, USA) and GraphPad Prism 5.0 (GraphPad Software Inc, La Jolla, CA, USA). Responses were normalized to the maximal response elicited by GABA. All data were presented as mean ± SEM, comparisons between groups were analyzed using one-way analysis of variance (ANOVA) with Dunnett's post hoc test or unpaired t-test, and significant differences were considered if p < 0.05.