A photoswitchable GABA receptor channel blocker.

BACKGROUND AND
PURPOSE: Anion-selective Cys-loop receptors (GABA and glycine receptors) provide the main inhibitory drive in the CNS. Both types of receptor operate via chloride-selective ion channels, though with different kinetics, pharmacological profiles, and localization. Disequilibrium in their function leads to a variety of disorders, which are often treated with allosteric modulators. The few available GABA and glycine receptor channel blockers effectively suppress inhibitory currents in neurons, but their systemic administration is highly toxic. With the aim of developing an efficient light-controllable modulator of GABA receptors, we constructed azobenzene-nitrazepam (Azo-NZ1), which is composed of a nitrazepam moiety merged to an azobenzene photoisomerizable group. EXPERIMENTAL APPROACH: The experiments were carried out on cultured cells expressing Cys-loop receptors of known subunit composition and in brain slices using patch-clamp. Site-directed mutagenesis and molecular modelling approaches were applied to evaluate the mechanism of action of Azo-NZ1. KEY
RESULTS: At visible light, being in trans-configuration, Azo-NZ1 blocked heteromeric α1/β2/γ2 GABAA receptors, ρ2 GABAA (GABAC ), and α2 glycine receptors, whereas switching the compound into cis-state by UV illumination restored the activity. Azo-NZ1 successfully photomodulated GABAergic currents recorded from dentate gyrus neurons. We demonstrated that in trans-configuration, Azo-NZ1 blocks the Cl-selective ion pore of GABA receptors interacting mainly with the 2' level of the TM2 region. CONCLUSIONS AND IMPLICATIONS: Azo-NZ1 is a soluble light-driven Cl-channel blocker, which allows photo-modulation of the activity induced by anion-selective Cys-loop receptors. Azo-NZ1 is able to control GABAergic postsynaptic currents and provides new opportunities to study inhibitory neurotransmission using patterned illumination.

azobenzene‐nitrazepam

transmembrane domain

holding membrane potential

What is already known

Pharmacological activity of ligands can be regulated with azobenzene photochromic group, which changes its configuration upon illumination with light of different wavelength.

What this study adds

Azobenzene‐nitrazepam based photochromic compound Azo‐NZ1 is a light controllable channel blocker of heteromeric GABAA, homomeric GABAC ρ2 receptors, and α2 GlyRs.

Azo‐NZ1 modulates in a light-dependent manner synaptic GABAergic currents in dentate gyrus of hippocampal brain slices.

What is the clinical significance

Azo‐NZ1 modulates in a light‐dependent manner synaptic GABAergic currents in dentate gyrus of hippocampal brain slices.

INTRODUCTION

The major inhibitory neurotransmission in the mammalian CNS is provided by GABA, (also known as GABAC ρ receptor; see Alexander et al., 2015), and glycine receptors (Langosch, Becker, & Betz, 1990). In different combinations, these receptors are widely expressed throughout the nervous system and are involved in physiological functions such as cognition, learning, memory, and motor control. Disequilibrium in the functioning of GABA and glycine (Gly) receptors leads to important disorders including epilepsy, anxiety, depression, misbalance of blood pressure, sleep disorders, schizophrenia, and hyperekplexia (Johnston, Chebib, Hanrahan, & Mewett, 2003; Tan, Rudolph, & Lüscher, 2011; Sieghart, 2015; Lynch, 2004).

The activity of GABAA receptors can be successfully regulated with allosteric modulators like benzodiazepines, including flumazenil and diazepam. They are among the most popular pharmacological drugs and are commonly prescribed to treat insomnia, anxiety, and convulsions. Benzodiazepines can potentiate GABA receptors through interaction with different sites in the receptor displaying nano‐ and micromolar affinities (Sieghart, 2015). Therefore, benzodiazepine derivatives are convenient scaffolds to develop new modulators of GABA receptor function.

Light‐switchable ligands have become a useful tool to specifically control ionic channels, cellular functions, the activity of neuronal circuits, and behaviour (Gorostiza & Isacoff, 2008; Kramer, Mourot, & Adesnik, 2013). The pharmacological activity of ligands can be regulated with photochromic groups—molecules that change their configuration upon illumination with different wavelengths. Azobenzene undergoes a rapid trans–cis photoisomerization and has been widely used for this purpose. In the dark, or under visible light irradiation, azobenzene is in the elongated trans‐configuration, while upon illumination with UV light (360–380 nm), it isomerizes to its cis‐state and becomes shorter by several angstroms (Gorostiza & Isacoff, 2008; Yu & Ikeda, 2004).

With the aim of regulating the activity of GABA receptors with light, we constructed a series of azobenzene derivatives of nitrazepam and characterized their properties. Herein, we describe the interaction of one of these compounds, azobenzene‐nitrazepam‐1 (Azo‐NZ1), with several subtypes of anion‐selective Cys‐loop receptors (GABAA, GABAC and glycine receptors). Azo‐NZ1 successfully photocontrolled GABAergic currents in cultured cells and in brain slices, but in contrast to the intended design to obtain an allosteric ligand of the classical diazepam site, we found that the azobenzene sulfonyl group introduced alters the binding properties of the compound, enabling light‐controlled blocking of the chloride‐selective ion pore. Thus, Azo‐NZ1 reveals a class of photochromic ligand with a structure capable of blocking the anion‐selective channel pore of several Cys‐loop receptors.

METHODS

Synthesis of compounds

7‐Aminonitrazepam (Guandalini et al., 2008; Severino et al., 2008) and tetrabutylammonium 4‐nitrosobenzenesulfonate (Priewisch & Rück‐Braun, 2005) were synthesized according to previously reported procedures. Commercial reagents and starting materials were purchased from Acros Organics, Alfa‐Aesar, Fisher Scientific, Sigma Aldrich, or VWR and used without any further purification. Solvents used were of p.a. quality and dried according to common procedures, if necessary. Commercially obtained phosphate buffer (pH = 7.4) was used for investigations of the photochromic properties. Dry nitrogen was used as inert gas atmosphere. A Biotage Isolera flash purification system with UV/Vis detector using Sigma Aldrich MN silica gel 60 M (40–63 μm, 230–400 mesh) for normal phase or pre‐packed Biotage SNAP cartridges (KP C18 HS) for reversed phase chromatography was used for automated flash column chromatography. Reaction monitoring via TLC and determination of R values was accomplished on alumina plates coated with silica gel (Merck silica gel 60 F254, 0.2 mm). Melting points were measured with a Stanford Research Systems OptiMelt MPA 100 device and are uncorrected. NMR spectra were measured on Bruker Avance 300 (1H 300.13 MHz, 13C 75.48 MHz), Bruker Avance 400 (1H 400.13 MHz, 13C 100.61 MHz), and Bruker Avance III 600 (1H 600.25 MHz, 13C 150.95 MHz) instruments. The spectra are referenced against the NMR solvent (DMSO‐d 6: δH = 2.50 ppm), and chemical shifts, δ, are reported in ppm. Resonance multiplicity is abbreviated as follows: s (singlet), d (doublet), t (triplet), q (quartet), m (multiplet), and b (broad). Carbon NMR signals are reported using DEPT 135 and 1H–13C HSQC spectra with (+) for primary/tertiary, (−) for secondary, and (q) for quaternary carbons. An Agilent Q‐TOF 6540 UHD (ESI–MS) instrument was used for recording mass spectra. UV/Vis absorption spectroscopy was accomplished using a Varian Cary BIO 50 UV/Vis spectrometer in 10‐mm quartz cuvettes. IR spectra were recorded on an Agilent Cary 630 FT‐IR spectrometer, and the peak positions are reported in wavenumbers (cm−1). Analytical HPLC measurements were performed on an Agilent 1220 Infinity LC (column: Phenomenex Luna 3 μM C18(2) 100 Å, 150 × 2.00 mm; flow: 0.3 ml·min−1 at 30°C; solvent A: MilliQ water with 0.05% vol TFA; solvent B: MeCN). The ratios in the photostationary states (PSSs) were determined by HPLC measurements with a detection wavelength at the isosbestic points. For determination of the thermal half‐lives, the solutions were irradiated until the photostationary state was reached. Then the solutions were left at 25°C, and the recovery of the absorbance of the trans‐isomer at λmax was measured. Consequently, the thermal half‐life was calculated by fitting the data with a single exponential function. An Agilent 1260 system (column: Phenomenex Luna 10 μM C18(2) 100 Å, 250 × 21.2 mm; flow: 22.0 ml·min−1; solvent A: MilliQ water with 0.05% vol TFA; solvent B: MeCN) was used for preparative HPLC purification. Light sources for irradiation: λ = 365 nm (Herolab handheld lamp UV‐6 L, 6 W), λ = 455 nm (OSRAM Oslon SSL 80 LED, 700 mA, 1.12 W). The power of the light is given based on the specifications supplied by the company when the lamps were purchased. Final compounds for biological testing possess a purity ≥95%, as determined by HPLC measurements with detection at 220 and 254 nm.

Azo‐NZ1 ((E)‐4‐((2‐oxo‐5‐phenyl‐2,3‐dihydro‐1H‐benzo[e][1,4]diazepin‐7‐yl)diazenyl)benzenesulfonic acid) was synthesized according to the following procedure (Figure 1b). Tetrabutylammonium 4‐nitrosobenzenesulfonate (236 mg, 0.54 mmol, 1.0 eq.) was added to a solution of 7‐aminonitrazepam (68 mg, 0.27 mmol, 1.0 eq.) in acetic acid (2 ml) and CH2Cl2 (1 ml). After the mixture had been stirred for 24 hr at room temperature, the solvent was removed in vacuo. Purification by automated flash column chromatography (CH2Cl2/MeOH, 3–25% MeOH) and subsequent preparative HPLC (2–65% MeCN in 10 min, t R = 7.2 min) yielded Azo‐NZ1 (69 mg, 61%) as a yellow solid. R 0.03 (CH2Cl2/MeOH 9:1); mp 280°C (decomposition); 1H‐NMR (600 MHz, DMSO‐d 6) δ = 11.32 (s, 1H), 8.23 (dd, J = 8.8, 2.4 Hz, 1H), 7.81–7.78 (m, 2H), 7.77–7.73 (m, 3H), 7.68 (t, J = 7.4 Hz, 1H), 7.64 (dd, J = 8.3, 1.3 Hz, 2H), 7.57 (t, J = 7.7 Hz, 2H), 7.52 (d, J = 8.9 Hz, 1H), 4.34 (s, 2H); 13C‐NMR (151 MHz, DMSO‐d 6) δ = 172.5 (q), 168.9 (q), 151.4 (q), 151.1 (q), 146.5 (q), 143.0 (q), 135.8 (q), 132.5 (+), 130.7 (+), 128.8 (+), 128.6 (+), 126.8 (+), 126.2 (+), 124.4 (q), 122.9 (+), 122.2 (+), 54.6 (−); IR (neat) ν = 3489, 3135, 2930, 1715, 1614, 1484, 1435, 1387, 1342, 1230, 1163, 1115, 1029, 1006, 846, 742, 697 cm−1; HRMS (ESI) calculated for C22H16N4O4S (M + H)+ m/z = 421.0965; found 421.0964.

Figure 1

Design of Azo‐NZ1. (a) Chemical formula of diazepam and nitrazepam. (b) Synthetic approach for the synthesis of Azo‐NZ1. (c) UV–Vis spectra of azo‐nitrazepam Azo‐NZ1 (50 μM in phosphate buffer +0.1% DMSO) from the trans‐isomer (black), the PSS at irradiation with UV light of λ = 365 nm (purple), and the PSS at irradiation with blue light of λ = 455 nm (blue). (d) Schematic representation of Azo‐NZ1 in its trans‐ and cis‐configurations

Cell culture and transfection

The experiments were carried out on cultured CHO cells (CHO‐K1, ATCC Cat# CCL‐61, RRID:CVCL_0214) obtained from the American Type Tissue Culture Collection (ATCC, Molsheim, France) that were maintained in culture conditions as previously described (Maleeva, Peiretti, Zhorov, & Bregestovski, 2017; Mukhtarov et al., 2013). For electrophysiological analysis, cells were transfected with cDNAs of several Cys‐loop receptors. For expression of heteromeric GABAA receptors, cells were simultaneously transfected with cDNA (concentrations 0.9–1.2 μg·μl−1) of α1, β2, and γ2 (isoform S) subunits. For expression of homomeric GABAC receptors, cells were transfected with cDNA of ρ1 or ρ2 subunits. For the expression of homomeric glycine receptors and 5‐HT3A receptors, cDNAs of α1, α2, and A subunits, respectively, were used.

One day before transfection, cells were plated on the cover slips (12 mm in diameter) and placed inside 35‐mm cell culture dishes with 2 ml of medium. Transfection was performed using the Lipofectamine 3000 protocol (Life Technology, USA). To facilitate identification of transfected cells, a GFP (0.5 μg·μl−1) was added to the transfection mixture. Electrophysiological recordings were performed in the fluorescent cells 24–72 hr after transfection.

Electrophysiological recordings on CHO cells

Whole‐cell recordings were performed at room temperature (20–25°C) using an EPC‐9 amplifier (HEKA Elektronik, Germany). Cells were continuously superfused with external solution containing (mM): NaCl 140, CaCl2 2, KCl 2.8, MgCl2 4, HEPES 20, glucose 10; pH 7.4; 320–330 mOsm. Intracellular solution used for filling recording patch pipettes contained (mM): KCl 140, MgCl2 2, MgATP 2, HEPES10, BAPTA (tetrapotassium salt) 2; pH 7.3; 290 mOsm.

Recording pipettes were pulled from borosilicate glass capillaries (Harvard Apparatus Ltd, USA) and had resistances of 5–10 MOhms. For the rapid replacement of the solutions, the fast application system was used. Three parallel rectangular tubes (100 × 100 μm) were positioned 40–50 μm above the recorded cell. The movement of the tubes was controlled by a computer‐driven fast exchange system (SF 77A Perfusion Fast‐Step, Warner, USA) allowing a 10–90% solution exchange in 3–5 ms, as measured by open electrode controls (1/10 external solution/water). Cells with a low input resistance (<150 MOhms) and a rapid run‐down (>30% with repetitive application) were excluded from analysis.

The effect of Azo‐NZ1 was examined either using a “long” protocol for application of solutions, at constant holding membrane potentials (Vh) of +30 and −30 mV or using a “ramp” protocol that allowed a linear change of membrane potential from −60 to +60 mV (see Figure 3c). For GABAC ρ2, ρ1 P2′S, and α2 glycine receptors, the controls correspond to the amplitude of the current just before application of Azo‐NZ1, whereas for GABAA receptors, the control corresponds to the amplitude of the current at wash‐out, due to the desensitization of this type of receptor.

Figure 3

Strength of suppression of GABAA receptors by Azo‐NZ1 depends on the membrane potential and becomes stronger at higher agonist concentration. (a) Representative trace of the current induced by application of 5‐μM GABA and by co‐application of 5‐μM GABA with 50‐μM Azo‐NZ1, Vhold = −30 mV. (b) Representative recording of the current induced by saturating concentration of GABA (300 μM) and by a mixture of 300‐μM GABA with 50‐μM Azo‐NZ1, Vhold = −30 mV. (c) Cumulative current‐voltage relationship curves obtained during application of 5‐μM GABA, a mixture of 5‐μM GABA with 50‐μM Azo‐NZ1 at visible light, and GABA + Azo‐NZ1 upon UV illumination (n = 5). Scheme of the ramp protocol is presented in the inset on the left. (d) Summary of GABAA receptor‐mediated suppression induced by Azo‐NZ1. Mean relative amplitude (%) of currents induced by a mixture of Azo‐NZ1 with low (5 μM) and saturating (300 μM) concentrations of GABA at visible light and upon UV illumination. Grey column: 5‐μM GABA +50‐μM Azo‐NZ1 at visible light; yellow column: 300‐μM GABA +50‐μM Azo‐NZ1 at visible light; violet columns: 5‐ or 300‐μM GABA +50‐μM Azo‐NZ1 at UV illumination. Data are presented as mean ± SEM; Vhold = −30 mV summary from six to eight cells, *P < .05

UV (365 nm) LED (Thorlabs) was placed at the distance of 4–5 cm from the cell that was being receorded. The power of UV light was reaching 0.6 mW·mm−2 at the level of recording chamber, as determined using an optical power metre (Thorlabs).

Electrophysiological recording on brain slices and electrical stimulation

Experiments were performed on white laboratory ICR outbred mice of both genders, kept at 12 hr “day”/“night” cycle. Other details are described previously (Petukhova, Ponomareva, Mukhamedyarov, Maleeva, & Bregestovski, 2018). All animal protocols and experimental procedures were approved by the Local Ethics Committee of Kazan State Medical University (N742.13.11.84 and N1045‐72) and by the INSERM Ethics Committee for Animal Experimentation (#30‐03102012). Animal studies are reported in compliance with the ARRIVE guidelines (Kilkenny et al., 2010) and with the recommendations made by the British Journal of Pharmacology.

Coronal slices, containing hippocampus, were obtained from 3‐ to 4‐week‐old mice. Animals were decapitated under isoflurane anaesthesia. Whole brains were rapidly removed from the skull and immersed for 10 min in ice‐cold artificial CSF (aCSF) solution containing (in mM): NaCl 126, KCl 3.5, CaCl2 2, MgCl2 1.3, NaHPO4 1.2, glucose 10, NaHCO3 25 (pH 7.3–7.4, 290–300 mOsm). The aCSF was continuously oxygenated with 95% O2 and 5% CO2 to maintain the physiological pH. Sagittal slices (350 mm) were cut using a tissue slicer (Microm International, Germany) in high K+ solution, containing (in mM): Kgluconate 120, HEPES‐acid 10, Nagluconate 15, EGTA 0.2, NaCl 4 (pH 7.2, 290–300 mOsm). After being cut, slices were placed for 15 min at room temperature in a choline‐based solution, containing (in mM): choline chloride 110, KCl 2.5, NaH2PO4 1.25, MgCl2 10, CaCl2 0.5, NaHCO3 25, glucose 10, sodium pyruvate 5 (pH 7.3–7.4, 290–300 mOsm). Then, slices were placed in a chamber filled with oxygenated aCSF. Before use, slices were allowed to recover for at least 1 hr at room temperature. Experiments were performed during the period of 1–6 hr after slicing.

During the recordings, slices were placed in a conventional chamber superfused with aCSF (32–34°C). The inhibitory postsynaptic currents evoked (eIPSCs) were recorded from granule cells of dentate gyrus using the patch‐clamp technique in whole‐cell configuration at −70 or 0 mV. Patch electrodes were pulled from borosilicate capillary glass (Sutter Instruments flaming/brown micropipette puller P‐97) and filled with intracellular solution containing (in mM): Kgluconate 120, KCl 20, MgCl2 2, MgATP 2, HEPES/KOH 10, BAPTA 1; pH 7.3; 290 mOsm. The resistance of electrodes used was 5–7 MΩ. Membrane currents were recorded at 3–10 kHz using an EPC‐9 patch‐clamp amplifier (HEKA Elektronik, Germany).

In granule cells of dentate gyrus, postsynaptic currents consist of two main populations: glutamatergic and GABAergic. In our experimental conditions, the reversal potentials for glutamate‐ and GABA‐induced currents were about 0 and −50 mV, respectively. To isolate GABAergic synaptic events from glutamatergic ones without using specific pharmacological antagonists, we performed analysis of Azo‐NZ1 action on eIPSCs at Vh = 0 mV. At this potential, glutamatergic events are not detectable, and GABAergic currents appear as outwardly directed deflections. An application of bicuculline at the end of recordings resulted in the complete suppression of these currents, ensuring its pure GABAergic nature.

For the induction of GABAergic responses, stimulation pulses were generated by the DS3 Constant Current Isolated Stimulator (Digitimer, England). Pulses were delivered every 10 s at Vh = 0 mV, the amplitudes of control eIPSCs ranged from 50 to 400 pA. Electrodes for stimulation were prepared from double‐barrelled borosilicate theta‐tubes (OD = 1.5 mm, ID = 1.0 mm, septum = 0.2 mm, from Warner Instruments, USA). They were pulled using the programme similar to those for preparing patch‐clamp recording pipettes, polished, and backfilled with aCSF. To deliver an electrical stimulus, two fine silver wires were inserted into the theta‐glass.

After a whole‐cell configuration had been obtained with a neuron, the stimulation electrode was placed on the granule cell layer in close proximity within the recording cell (distance 100–200 μm) located towards the hilar region. This allowed the stimulation of presynaptic fibres from the presumed GABAergic basket cells, as described by others (Haselmann, Röpke, Werner, Kunze, & Geis, 2015; Korshoej, Holm, Jensen, & Lambert, 2010).

Previous studies demonstrated that azobenzene‐based UV‐controlled molecules might be successfully used in living tissues for research purposes (Lin et al., 2015; Yue et al., 2012). However, after prolonged exposure, UV and near‐UV irradiation can be damaging (Kienzler et al., 2013). To avoid direct illumination of brain slices with UV light, which might modify synaptic function, we illuminated the aCSF with diodes emitting either UV (365 nm) or blue (455 nm) light. The speed of the perfusion was about 20 ml·min−1, and it was taking 18‐20 s for the illuminated solution to arrive to the experimental chamber.

Effects of Azo‐NZ1 on the amplitudes of eIPSCs were studied using the following protocol: 20–30 eIPSCs were recorded in control condition, then slices were incubated in aCSF + Azo‐NZ1 for 5–10 min at visible or blue (455 nm) light (the compound in trans‐configuration); the aCSF + Azo‐DZ1 was illuminated with UV light for 3–5 min (switching the compound into cis‐configuration); and subsequently illuminated for 3–5 min with blue light. Changes in UV/blue light illumination could be repeated several times on the same neuron before it is washed with control aCSF. Control tests showed that UV illumination of the pure aCSF for 5 min had no effect on eIPSCs.

Data and statistical analysis

The data and statistical analysis comply with the recommendations on experimental design and analysis in pharmacology. Electrophysiological recordings were performed using PatchMaster (HEKA Electronic, Germany) software. To plot concentration–response curves, responses to different concentrations of GABA were fitted using a non‐linear fitting routine of the Origin 7.5 software (OriginLabs, USA) with the Hill equation: where I is the normalized current amplitude induced by the agonist at concentration [A], I max is a maximal current induced at given cell, n is the Hill coefficient, and EC50 is the concentration at which a half‐maximum response was induced.

Ionic current recordings were analysed with Igor Pro 6.02 and Origin 9.0 software. In some cases, data were normalized to perform a comparison between different types of receptor or different agonist concentrations. Data are presented as means ± SEM. Significance of difference was evaluated by using two‐sample t test, paired sample Wilcoxon signed rank test, and Mann–Whitney U test. Differences were considered significant at a value of P below .05.

Materials

All the drugs were obtained from Tocris or Sigma–Aldrich (France). Stock solutions of Azo‐NZ1 (10 mM) and diazepam (50 mM) were prepared using DMSO and then diluted to the final concentration in extracellular solution. Stock solutions of GABA (1 M), glycine (1 M), 5‐HT (10 mM), APV (40 mM), and bicuculline (10 mM) were prepared using MilliQ water.

Modelling

Five different GABA receptor structures were used, all corresponding to open channel states. For GABAA receptors, we used the homology model of the heteropentameric α1/β2/γ2 receptor from reference (Bergmann, Kongsbak, Sørensen, Sander, & Balle, 2013). Two homology models were generated for the ρ1 and ρ2 GABAA (GABAC ) receptors, using as a template the GluCl receptor in the open channel state (PDB code 3RI5; Hibbs & Gouaux, 2011) and the Swiss‐Model server (Biasini et al., 2014). The sequence identity between the GluCl template and the ρ1 and ρ2 receptors is 37% and 36.6%, and it is even higher when only the pore‐lining helices are considered (54.2% and 58.3%, respectively). Indeed, GluCl has been proposed to be the best template to model the structure of GABAA ρ receptors (Naffaa, Chebib, Hibbs, & Hanrahan, 2015). For the S2′G ρ1 and P2′S ρ2 mutant receptors, the mutation was introduced in silico using the Molefacture tool (version 1.3) in VMD (version 1.9.2; Humphrey, Dalke, & Schulten, 1996).

While this manuscript was under review, new cryo‐electron microscopy structures of GABAA receptors were published (Laverty et al., 2019; Masiulis et al., 2019). These structures correspond to either closed or desensitized states of the receptor and hence were not used for our docking analysis Nonetheless, we superimposed the structural model of the open state used here (Bergmann, Kongsbak, Sørensen, Sander, & Balle, 2013) with the cryo‐EM structures of GABAA receptor in the closed and desensitized states (Figure S9), in order to investigate the possibility of Azo‐NZ1 binding to these other two pore conformations. In the closed state, the pore is narrower than in the open state (Figure S9A,B) and thus would hinder binding of the longer and bulkier trans‐Azo‐NZ1, compared to the smaller pore blocker picrotoxinin captured in the cryo‐EM structure. For the desensitized state, the pore size is not as different from the open state (Figure S9B,C), and thus, we cannot exclude binding of trans‐Azo‐NZ1. However, we speculate that the non‐chloride‐conductive desensitized state might be less suitable to bind the chloride‐mimic sulphonate group of trans‐Azo‐NZ1. In addition, binding of trans‐Azo‐NZ1 to Ser2′ could prevent desensitization, similarly to picrotoxinin (Gielen, Thomas, & Smart, 2015) and thus would be incompatible with a desensitized state.

Azo‐NZ1 ligand

The initial structures of the Azo‐NZ1 compound (cis‐ and trans‐isomers) were created employing Avogadro (version 1.1.1; Hanwell et al., 2012). For each isomer, two 1,4‐diazepine ring conformations, M and P, were considered, which differ in orientation (below or above the plane, respectively) of C3 and the phenyl substituent of C5 (Figure S10). For classical benzodiazepines that bind to the canonical allosteric site, the M conformation is the bioactive one (i.e., with higher affinity for receptor; Richter et al., 2012). However, it is not known a priori whether Azo‐NZ1 would exhibit similar conformational preferences, since it binds in a distinct site. Therefore, all four ligand structures (cis/M, cis/P, trans/M, and trans/P) were optimized using quantum mechanical calculations in order to (a) optimize the manually built ligand structures and (b) estimate the relative stability of the M and P conformers. Calculations were performed with the Gaussian 09 (G09) programme package (Frisch et al., 2013), using density functional theory (Rajagopal & Callaway, 1973), with the B3LYP functional (Stephens, Devlin, Chabalowski, & Frisch, 1994) and the 6–31++G(d,p) basis set. For the trans‐isomer, the two conformers differ only by 0.03 kcal·mol−1, and thus, their Boltzmann populations are very similar (51.4% and 48.6% for P and M, respectively). In other words, the two conformers can be present at room temperature. In the case of the cis‐isomer, the P conformer is significantly more stable (by 3.5 kcal·mol−1) than the M one, and thus, it is the predominant conformer (with a Boltzmann population of 99.7%).

Docking calculations

Autodock Vina (version 1.1.2; Trott & Olson, 2010) was employed for ligand–receptor docking. Given that experimental data indicate that Azo‐NZ1 binds inside the pore, we centred our search space around the M2 helixes of the pore. A flexible docking approach was used, in which both the receptor and the ligand were considered as flexible. On the one hand, the ligand was allowed to change its geometry (by exploring all the possible ligand torsions) in order to optimize its fit inside the pore. On the other hand, the side chains of the pore‐lining residues at positions 2′, 6′, 9′, 13′, and 16′ were allowed to move and adapt to the ligand poses. The maximum energy difference between the best and worst binding modes and the exhaustiveness were set to default values (3 and 8 kcal·mol−1, respectively). Instead, the maximum number of modes was increase to 20 in order to increase the docking sampling. This protocol was repeated 10 times [for GABAA ρ (GABAC) receptors] and eight times (for GABAA receptors), starting with different random seeds, so that a total number of 200 and 160 binding modes was obtained for each of the four possible conformers of azo‐NZ1 (200 for cis/M and cis/P, and 160 for trans/M and trans/P). The docking poses of each conformer were analysed separately. For simplicity, only the trans/P and cis/P results are discussed in the main text. The results of the corresponding M conformers are almost identical (and in the case of cis/M, it has very low population).

We would like to note here that, upon UV light irradiation, cis‐Azo‐NZ1 may either remain in the channel (assuming another binding conformation that unblocks the pore) or exit the pore (reaching the bath solution). The outcome will depend on the mean open time of the channel and the dissociation rate of cis‐Azo‐NZ1. Unfortunately, the dissociation rate is particularly difficult to study either experimentally or computationally. Nevertheless, we expected it to be slow, based on the observations for GABAA receptors with another pore blocker picrotoxinin (Gielen, Thomas, & Smart, 2015; Korshoej, Holm, Jensen, & Lambert, 2010). Therefore, since with the data at hand we cannot distinguish between the two possibilities put forward above, we decided to perform docking calculations not only with trans‐Azo‐NZ1 but also with the cis‐isomer, in order to model the situation immediately after UV irradiation. These dockings are aimed at investigating the change in the receptor–ligand binding mode upon trans–cis isomerization (both in terms of position along the pore and interactions with the pore‐lining residues).

Analysis of the docking results

We used two different approaches to pinpoint the exact binding site of Azo‐NZ1. The first approach was to use the number density of the sulfonate group to follow ligand binding. Previous studies have successfully used this type of analysis to identify ligand binding sites in other ion channels (Raju, Barber, LeBard, Klein, & Carnevale, 2013). The underlying assumption is that regions of continuous density (or high occupancy) should represent regions of tighter binding. We computed the number density value, using the Volmap plugin (Cohen, Arkhipov, Braun, & Schulten, 2006) of VMD (Humphrey, Dalke, & Schulten, 1996). For each binding pose, the sulfonate group of Azo‐NZ1 was considered a “particle” by replacing each of its atoms with a normalized Gaussian distribution (with width 1.5 times its atomic radius) and additively distributing those Gaussians on a three‐dimensional grid (with dimensions 0.5 × 0.5 × 0.5 Å3). The resulting particle density is thus equivalent to the number of binding poses whose sulphonate group is located in a given volume of the pore.

The second approach was based on the analysis of the receptor–ligand interactions. First, the ligand binding poses were clustered using the quality threshold algorithm implemented in VMD (https://github.com/luisico/clustering), and then the representative structure of the most populated cluster(s) was analysed using the Binana algorithm (Durrant & McCammon, 2011). The images of the modelling section were generated with UCSF Chimera (Pettersen et al., 2004) or VMD (Humphrey, Dalke, & Schulten, 1996).

Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY (Harding et al., 2018), and are permanently archived in the Concise Guide to PHARMACOLOGY 2017/18 (Alexander et al., 2017).

RESULTS

Design of Azo‐NZ1

The design of our photochromic ligand was based on the diazepam structure (Figure 1a), a widely used modulator of GABAergic function. Benzodiazepines are capable of binding in several interaction sites in the receptor (Middendorp, Maldifassi, Baur, & Sigel, 2015; Sieghart, 2015), and this offers, in principle, greater flexibility to introduce relatively bulky substitutions like azobenzene without leading to a loss of functional activity of the compound. This choice was strengthened by the fact that the benzodiazepine nitrazepam, a diazepam derivative (Figure 1a), can be easily functionalized for a subsequent Mills reaction via reduction of its nitro‐group (Severino et al., 2008) and thus be directly incorporated as part of the azobenzene (Figure 1b).

Therefore, the nitro‐group of nitrazepam was reduced providing an aniline, which was used for further functionalization to set up the azo‐bridge in a Mills‐type reaction (see Section 2). To increase the solubility of the azo‐nitrazepam, sulfanilic acid was used as precursor for the formation of the nitroso compound (Priewisch & Rück‐Braun, 2005) providing a negative charge at physiological pH values (Kim, Gao, & Burgess, 2009). This compound showed robust photochromism with a preference for the trans‐configuration in darkness and under visible light (455 nm), while under UV illumination (365 nm), the cis‐configuration is favoured (Figures 1c,d and S11).

Action of Azo‐NZ1 on heteromeric GABA α1/β2/γ2 receptors

Analysis of the GABA dose/response curves showed that EC50 for α1/β2/γ2 GABAA receptors was 8 ± 3 μM (Figure 2a; n = 6), which was close to the previously reported values (Fisher, 2004). Diazepam (10 μM) co‐applied with non‐saturating concentration of GABA caused an increase in the current amplitude (Figure 2b), which is in accordance with previous observations (Bormann & Clapham, 1985). UV illumination neither modified the currents induced by GABA nor by GABA plus diazepam (Figure 2b).

Figure 2

Azo‐NZ1 modulates activity of GABAA α1/β2/γ2 receptors in a UV‐dependent manner. (a) Cumulative dose/response curve for GABA‐induced currents in control conditions (n = 6). (b) Traces of currents induced by 5‐μM GABA (left panel) and by co‐application of 5‐μM GABA with 10‐μM diazepam (right panel) at Vh = −30 mV. Durations of GABA and GABA/diazepam applications are indicated by black bars above the traces. Duration of UV illumination is indicated by violet rectangles. Note the absence of the effect of UV illumination on the responses. (c) Effect of 70‐μM Azo‐NZ1 on the amplitude of currents induced by 5‐μM GABA at Vh = +30 mV (top trace) and −30 mV (bottom trace). Note the stronger inhibition of currents by Azo‐NZ1 at positive potentials. (d) Cumulative dose‐response curve for Azo‐NZ1 at GABAA receptors (n = 6)

Ionic currents were evoked by 5 μM of GABA alone or with addition of Azo‐NZ1. The photochromic compound in concentrations 0.5–1 μM did not produce effect on the GABA‐induced currents, while application of Azo‐NZ1 at 10 μM and higher concentrations caused its decreasing. As illustrated in Figure 2c, Azo‐NZ1 at 70 μM produced a prominent decrease of the amplitude of GABAA‐mediated currents under visible light, when the compound was in trans‐configuration. Illumination with UV light (365 nm) resulted in recovery of currents till values were close to the control. The degree of Azo‐NZ1‐induced suppression at positive potential was more pronounced: the amplitude of GABA‐induced currents decreased by 78% and 51% at +30 mV and −30 mV, respectively (Figure 2c). Analysis of concentration dependencies of Azo‐NZ1 action shows that at −30 mV, the IC50 was 67 ± 5 μM (n = 6; Figure 2d), and thus, further experiments were performed with concentrations of Azo‐NZ1 close to IC50.

Using a “ramp” protocol (Figure 3c, inset), we have shown that, under visible light, 50‐μM Azo‐NZ1 decreased currents induced by 5‐μM GABA by 76 ± 5% and 40 ± 4% at +60 mV and −60 mV, respectively (n = 5, P < .05), confirming that Azo‐NZ1 is more efficient at positive potentials (Figure 3c). The recovery during UV illumination was not complete (P < .05), presumably, because of incomplete photoswitching of Azo‐NZ1 in cis‐configuration.

The voltage dependence of the inhibitory action of Azo‐NZ1 suggested that the compound might act as an open channel blocker of GABAA receptors (Maleeva, Peiretti, Zhorov, & Bregestovski, 2017). To test this hypothesis, we compared the action of Azo‐NZ1 at different concentrations of the agonist. The amplitude of currents induced by GABA concentration below EC50 (5 μM) was decreased by application of Azo‐NZ1 (50 μM) (Figure 3a,d). Azo‐NZ1 co‐applied with saturating concentration of GABA (300 μM) produced an even stronger inhibitory effect (Figure 3b,d). Thus, Azo‐NZ1 acts as a non‐competitive antagonist of GABAA receptors, and its binding does not coincide with the agonist‐binding site. In both cases, UV illumination restores current amplitude, though not completely, close to the control values.

Action of Azo‐NZ1 on homomeric ρ1 and ρ2 GABAA (GABA) receptors

To further investigate the hypothesis that Azo‐NZ1 acts as a blocker of the GABA receptor channels without interacting with the diazepam‐binding site located at the α/γ subunits interface, we analysed the action of this photochromic ligand on benzodiazepine‐insensitive ρ1 and ρ2 GABAA (GABAC) homomeric receptors. Concentration‐dependency analyses showed that the EC50 for GABA of ρ1 and ρ2 GABAC receptors were similar (Figure 4a). In accordance with previous observations (Ragozzino et al., 1996), GABA‐induced currents mediated by these receptors do not exhibit desensitization and are characterized by a very long deactivation period after agonist application.

Figure 4

Azo‐NZ1 decreases currents via GABACρ ρ2 receptors in a UV‐dependent manner, while being not active at GABAC ρ1 receptors. (a) Cumulative dose‐response curves for GABA at GABAC ρ1 and GABAC ρ2 receptors. (b) Representative trace illustrating the absence of Azo‐NZ1 effect on GABAC ρ1 receptors at visible and UV light. (c) Representative trace of GABA ρ2‐mediated current induced by application of GABA 10 μM at visible light and upon UV illumination. Note the absence of UV effect on the amplitude of GABA‐induced current. (d) Representative traces of GABAC ρ2 currents induced by 10‐μM GABA and by a mixture of 10‐μM GABA with 100‐μM Azo‐NZ1 at +30 mV (upper panel) and at −30 mV (lower panel). (e) The relative amplitudes of GABAC ρ2‐mediated currents suppressed by Azo‐NZ1 (100 μM): at +30 and −30 mV in control (GABA 10 μM), after application of Azo‐NZ1 in visible light (orange column) and upon UV illumination (violet column). Mean ± SEM, n = 6–10, *P < .05. (f) Cumulative dose‐response curve for Azo‐NZ1 at GABA ρ2 receptors (n = 6)

We found that Azo‐NZ1 did not alter the activity of human ρ1 GABAC receptors (Figure 4b), but it enabled robust photoswitching of ρ2 receptors (Figure 4d). The IC50 for GABA ρ2 receptors comprised 128 ± 25 μM (n = 6), and thus, all the experiments with ρ2 receptors were performed at 100 μM of Azo‐NZ1, as this concentration was close to IC50 (Figure 4f). Under visible light, currents induced by saturating GABA concentration (10 μM) were decreased by 100‐μM Azo‐NZ1 at −30 and +30 mV, respectively (Figure 4d,e), that is, the inhibitory action of Azo‐NZ1 on ρ2 receptors is also voltage‐ dependent. UV illumination restored the amplitude of the current to the nearly control values. The amplitude of ρ2 currents induced by 10 μM of GABA without addition of Azo‐NZ1 was not affected by UV illumination (Figure 4c).

Molecular determinants of Azo‐NZ1 action

The well‐defined voltage dependence of the Azo‐NZ1 effect on GABAA and GABAC ρ2 receptors and its independence on the agonist concentration strongly suggest that Azo‐NZ1 in trans‐configuration blocks the ion pore of these receptors. It is well documented that the position 2′ of the pore‐forming transmembrane domain 2 (TM2) is crucial for the action of several pore‐blocking molecules in Cys‐loop receptors, including GABAC receptors (Xie, Song, Ripps, & Qian, 2008). An amino acid sequence alignment showed that TM2 domains of ρ1 and ρ2 subunits differ by the amino acids at the position 2′: ρ1 contains proline while ρ2 contains a serine residue (Figure 5a).

Figure 5

Effect of Azo‐NZ1 on mutant ρ1 and ρ2 GABAC receptors. (a) Amino acid sequence alignment of GABA receptors subunits TM2 Cl‐selective pore. Amino acids in 2′ position are highlighted in red. Note the high identity of the TM2 sequences between ρ1 and ρ2 subunits. (b) Representative traces of currents mediated by GABA ρ1 P2′S receptors suppressed by 100‐μM Azo‐NZ1 application at −30 and +30 mV. (c) Relative amplitude of GABA‐induced currents mediated by ρ1 P2′S receptors at +30 and −30 mV in control (10‐μM GABA), at application of 100‐μM Azo‐NZ1 in visible light (orange column) and upon UV illumination (violet column). Mean ± SEM, n = 7, *P < .05. (d) Representative trace depicting the absence of Azo‐NZ1 effect at ρ2 mutant S2′G receptors

To check the putative role of the 2′ residue in Azo‐NZ1 sensitivity, we substituted proline by serine at position 2′ of the ρ1 subunit. P2′S mutation did not affect the apparent affinity of the ρ1 receptor to GABA (EC50 = 0.98, n = 4) but made it sensitive to the action of Azo‐NZ1. Under visible light, 100‐μM Azo‐NZ1 inhibited currents induced by saturating concentrations of 10‐μM GABA (Figure 5b,c). UV illumination caused recovery of currents to 97 ± 1% and 90 ± 3%, respectively (Figure 5b).

Thus, we confirmed that Azo‐NZ1 in its trans‐configuration interacts with the pore of GABA receptors and identified the 2′ position of the TM2 domain as key determinant of the light‐dependent action of Azo‐NZ1. This conclusion was additionally confirmed in experiments demonstrating that the S2′G mutation in the ρ2 subunit influenced its interaction with Azo‐NZ1. Upon application of Azo‐NZ1 currents amplitude decreased by 14 ± 2% (Figure 5d, n = 5, P < .05), which might be due to weak interaction between Azo‐NZ1 and mutant receptor; importantly, UV effect was virtually absent. Results of the modelling analysis presented below are in line with our observations.

Interaction of Azo‐NZ1 with other members of Cys‐loop receptor family

GABA receptors and glycine receptors belong to the Cys‐loop receptors family and amino acid sequences forming their Cl‐selective pores are highly similar (Figure S1A). Thus, we asked if Azo‐NZ1 also acted as a glycine receptor channel blocker.

Trans‐Azo‐NZ1 produced only minor inhibitory effect in α1 glycine receptors—the current amplitude was reduced to 86 ± 3% (Figure S1B, n = 4, P < .05), with slightly stronger reduction upon illumination with UV light (to 73 ± 2%, P < .05). In contrast, in α2 glycine receptors, Azo‐NZ1 induced a prominent decrease of the current amplitude (to 23 ± 3%, n = 10, P < .01) that was altered by UV illumination (95 ± 7%, P < .01, n = 10; Figure S1B,C). The profile of α2 glycine receptor interaction with Azo‐NZ1 strongly resembles the one of GABAA and GABAC receptors, suggesting the same pore‐blocking mechanism of action.

In order to have a wider view of Azo‐NZ1 selectivity profile, we also tested it in a cation‐selective member of Cys‐loop receptor family, the 5HT3A receptor. In contrast to GABA receptors and glycine receptors, Azo‐NZ1 in trans or cis‐configurations did not interact in a specific manner with 5‐HT3A receptors (Figure S2, Supplementary Text). Thus, Azo‐NZ1 selectively interacts with anion‐permeable channels of the Cys‐loop receptor family, exhibiting a subunit specificity based on the nature of 2′ residues in TM2 pore.

Modelling of Azo‐NZ1 molecular interactions with GABA receptors

To get further insights into the mechanism of action of Azo‐NZ1 in GABA receptors, we performed molecular docking calculations. In view of the mutagenesis results indicating the relevance of the 2′ residues, the docking was carried out in the transmembrane region of the ion channel pore. Four different GABAC receptor structures were used: wild‐type ρ2, S2′G ρ2 mutant, wild‐type ρ1, and P2′S ρ1 mutant (see Section 2); all structures were modelled in the open channel state.

Docking of trans‐Azo‐NZ1 to the GABAC ρ2 receptor (Figure 6a) showed that the preferred binding site of the sulfonate group is located in the narrow 2′ position of the pore region (Figure S3A), at the centre of the ring formed by five serine residues (Figure 6b). Such a ligand orientation creates a strong hydrogen bond network involving the sulfonate oxygen atoms and the hydroxyl groups. In addition, a hydrogen bond between the hydroxyl group of one of the T13′ residues and the carbonyl of the nitrazepam core of Azo‐NZ1 holds the ligand in a linear orientation almost parallel to the pore axis. Further stabilization is provided by hydrophobic interactions involving the apolar region of the nitrazepam core, the side chain of L9′ and the methyl group of T6′ (Figure 6b). The length of the ligand (14 Å) is optimal to place its polar groups (sulfonate and amido) close to the polar side chains of S2′ and T13′, respectively, which are separated by 16 Å. In addition, the size of the sulfonate group is comparable with the pore diameter at the level of the 2′ serine ring, thus creating an obstacle that hinders the passage of the chloride ions through the pore both sterically and electrostatically (Figure S4). Interestingly, the negatively charged sulfonate group occupies the same position as the chloride ion trapped in X‐ray structures of GABAA receptor (Laverty et al., 2017) and glycine receptor (Du, Lü, Wu, Cheng, & Gouaux, 2015). Altogether, the architecture of trans‐Azo‐NZ1 fits perfectly in the pore region of the GABAC ρ2 receptor.

Figure 6

Results of the docking of Azo‐NZ1 in the transmembrane part domain of ρ2 GABAC receptor. (a) Tridimensional view of the ρ2 GABAC receptor. The S2′ and T13′ positions are marked with dashed lines. (b) Most likely docking pose of the trans‐isomer in the transmembrane domain. The interior of the pore is shown as viewed from the intracellular part. (c) Most likely pose for the cis‐isomer in the transmembrane domain. The interior of the pore is shown as viewed from the extracellular part. (d) Relative number of intermolecular interactions formed by the cis‐ and trans‐isomers of Azo‐NZ1 with residues of the transmembrane domain of ρ2 GABAC receptor

In contrast, the position of the sulfonate group of cis‐Azo‐NZ1 is more scattered all over the 2′‐13′ region of the pore (Figure S3B), and the number of poses with the sulfonate group in the 2′ region (i.e., the ones able to block the pore) is decreased by half. Analysis of the preferred binding pose of the ligand (Figure 6c) shows that, due to the staple‐like shape of cis‐Azo‐NZ1, the ligand is preferentially located further up in the pore and the number of interactions with the receptor is dramatically decreased compared to the trans‐isomer (Figure 6d). The ligand forms only one hydrogen bond with the pore residues, involving the tertiary amine of the benzodiazepine core and one hydroxyl group of T13′, and most importantly, the sulfonate group does not form any interaction with the pore‐lining residues, since it is positioned near L9′. The smaller number of interactions of the cis‐isomer compared to trans indicates that the ligand will have a higher probability of dissociation upon UV irradiation. Moreover, the ligand is mostly placed in the regions where the diameter of the pore is wider (around 11 Å, see Figure S4), and thus, it cannot block chloride conduction by either steric or electrostatic effects. Therefore, we predict that, even if cis‐Azo‐NZ1 remains bound, it would not decrease the channel current of the receptor. These results explain the higher capability of trans‐Azo‐NZ1 to block the pore of the GABAC ρ2 receptor compared to the cis‐isomer.

The importance of 2′ residues is also corroborated by the results obtained for the S2′G ρ2 mutant. In this case, the sulfonate group cannot form hydrogen bonds with the 2′ residues. As a consequence, the population of trans‐Azo‐NZ1 in the 2′ region decreases and the ligand is no longer able to block the pore (Figures S5B and S6A). Similar results were also obtained for the ρ1 receptor, which has proline at the 2′ region (Figures S6B and S7A). Conversely, mutation of proline to serine (P2′S) in the ρ1 receptor increases the population of the sulfonate in the 2′ region (Figure S7B) with respect to the WT receptor (Figure S7A). The associated binding mode of Azo‐NZ1 to P2′S ρ1 (Figure S6C) is very similar to the one for wild‐type ρ2 (Figure 6), consistent with the appearance of inhibition observed in our experiments (Figure 5).

We also performed docking calculations for the heteromeric α1/β2/γ2 GABAA receptor, using a model of the open channel state (Bergmann, Kongsbak, Sørensen, Sander, & Balle, 2013). For this receptor, the density analysis results are noisier due to the heterogeneous subunit composition (Figure S8A). Therefore, we can only say with certainty that the trans‐Azo‐NZ1 binding mode with the sulphonate bound at position 2′ is the most likely, similar to homomeric ρ2 GABAC receptor. Even though there is only one serine in this region (belonging to the γ2 subunit), the hydrogen bond interaction with the ligand is still present (Figure S8B). Moreover, although α1‐Val2′ and β2‐Ala2′ are not able to form hydrogen bonds, their volume is similar to that of Ser and thus still compatible with sulphonate binding at position 2′. In addition, the hydrogen bond between the carbonyl group of Azo‐NZ1 and T13′ is also maintained (Figure S8B). As a result, trans‐Azo‐NZ1 is clamped in the pore of the GABAA receptor (Figure S8C), as it was in the GABAC ρ2 receptor (Figure 6b) and thus is able to block chloride conduction.

Altogether, these observations demonstrate that Azo‐NZ1 is a photoswitchable blocker of GABAA and GABAC receptors, whose trans‐configuration interacts with residues located in the pore, in particular at positions 2′ and 13′. In trans‐Azo‐NZ1, the distance between the sulphonate group and the carbonyl group of the benzodiazepine core is optimal for Azo‐NZ1 to get clamped between these two residues, enabling pore blocking.

Effect of Azo‐NZ1 on synaptic currents in brain slices

Finally, to evaluate Azo‐NZ1 as photochromic modulator of GABAergic synaptic events, we performed electrophysiological recordings from neurons in dentate gyrus granular layer, which possess a strong GABAergic synaptic inputs (Coulter & Carlson, 2007; Haselmann, Röpke, Werner, Kunze, & Geis, 2015).

Whole‐cell recordings were performed at Vh = 0 mV and stimulation pulses induced by theta pipettes delivered at frequency 0.1 Hz, inducing a reliable GABAergic eIPSCs, which exhibited relatively low run‐down (about 20%) during long‐lasting recordings (>1 hr). In control conditions, amplitudes of eIPSCs varied in different neurons between 50 and 400 pA (n = 10).

For the neuron illustrated in the Figure 7a,b, amplitude of eIPSCs in control aCSF was 156.9 ± 4.5 pA, and it decreased to 100.6 ± 3.2 pA after addition of 100‐μM Azo‐NZ1. Switching to the solution pre‐illuminated with UV light (365 nm) caused an elevation of the amplitude, while the successive illumination with 455 nm resulted in its decrease. After washing with pure aCSF, the amplitude of eIPSCs recovered to near control value. Similar degrees of the light‐induced modulation was observed on five other neurons (Figure 7c). Application of 200‐μM Azo‐NZ1 caused much more pronounced inhibition (Figure 7d,e). On average, the suppression of eIPSCs amplitude by 100 and 200 μM of Azo‐NZ1 was to 58.6 ± 3.3% and 25.2 ± 6.0%, respectively (Figure 7d). Bicuculline at 10 μM caused complete inhibition of eIPSCs ensuring the GABAergic nature of the events (Figure 7e).

Figure 7

Effect of Azo‐NZ1 on the GABAergic postsynaptic currents in brain slices. (a) Traces of average evoked GABAergic postsynaptic currents (eIPSCs) recorded from dentate gyrus granule cells in the adult mice hippocampal slices: in the control conditions (Control); after addition to aCSF of 100‐μM Azo‐NZ1 (Azo‐NZ1); during perfusion of the slice with UV‐illuminated aCSF +100‐μM Azo‐NZ1 (Azo‐NZ1 + UV); during perfusion with blue light‐illuminated aCSF +100‐μM Azo‐NZ1 (Azo‐NZ1 + blue); 2–3 min of washing by aCSF (Wash). Each trace represents average of seven to 10 individual eIPSCs induced by stimulation with theta‐tube electrode placed in the granule cell layer. Vh = 0 mV. (b) Summary of the mean amplitudes of GABAergic eIPSCs recorded in control conditions; in the presence of 100‐μM Azo‐NZ1, at illumination either at 365 nm (purple column) or 455 nm (green column). Data from the experiment presented in (a). Whiskers area reflects the standard error range. *Significant difference with P < .05 (two‐sample t test, n = 7–10). (c) Mean percentages of the eIPSCs amplitude potentiation in the presence of 100‐μM Azo‐NZ1 by UV illumination. *Significant difference with P < .05 (paired sample Wilcoxon signed rank test, n = 5). (d) Summary of the percentages of the eIPSC amplitude decrease in the presence of 100‐ and 200‐μM Azo‐NZ1 at blue light illumination. */#Significant difference with P < .05 (paired sample Wilcoxon signed rank test and Mann–Whitney U test, respectively; n = 5). (e) Superimposed traces of eIPSCs illustrating decrease of currents at addition of 200‐μM Azo‐NZ1 and 10‐μM bicuculline. Note the complete suppression of events by bicuculline. Each trace represents average of six to 10 individual eIPSCs. Vh = 0 mV

These observations demonstrate that Azo‐NZ1 causes suppression of synaptic GABAergic currents in a light‐dependent manner.

DISCUSSION

Our study presents an azobenzene‐nitrazepam‐based photochromic compound (Azo‐NZ1) and characterization of its action on Cys‐loop receptors. Using electrophysiological, mutational, and molecular modelling analysis, we demonstrate that Azo‐NZ1 is a light‐controllable channel blocker of heteromeric GABAA, homomeric GABAC ρ2 receptors and α2 glycine receptors, heterologically expressed in cultured cells, as well as modulator of synaptic GABAergic currents in dentate gyrus neurons of hippocampal brain slices. Our attention was mostly focused on the action of Azo‐NZ1 at GABA receptors as they are the predominant subtype of inhibitory ligand‐gated receptors in higher brain regions (Bowery et al., 1984; Sieghart & Sperk, 2002).

Azo‐NZ1 as Cl‐channel blocker

Surprisingly, in our experiments, Azo‐NZ1 caused an inhibitory effect on the GABAA receptor, contrary to the potentiating effects via benzodiazepines binding sites. Moreover, similar inhibition was observed for GABA ρ2, which lacks the classical benzodiazepine site (Sieghart, 2015), and for α2 glycine receptors. While we cannot entirely rule out the allosteric interaction of Azo‐NZ1, the absence of modulation at nM concentrations and docking calculations (see Supporting Information) suggest that the functional effects of Azo‐NZ1 via the classical benzodiazepine site are negligible.

Our results strongly suggest that the photochrome acts as a blocker of GABA receptors Cl‐selective pore, interacting primarily with the 2′ position of the TM2 domain of the channel. The following results support the role of Azo‐NZ1 as GABA receptor pore blocker:

Azo‐NZ1‐induced suppression becomes stronger upon elevation of GABA concentrations;

the effect of Azo‐NZ1 is voltage dependent, being much stronger at positive membrane potentials;

mutation in the ρ1 GABAC receptor of a single amino acid situated at the 2′ level of the ion pore (P2′S) resulted in the appearance of blocking activity of Azo‐NZ1, with features similar to GABAC ρ2 receptors;

the S2′G mutation in ρ2 subunits resulted in the loss of the Azo‐NZ1 specific activity;

molecular docking calculations show that the length of trans‐Azo‐NZ1 fits in the pore region, with the sulfonate group blocking ion passage at the 2′ region.

In our experiments, we have achieved a robust photoswitching effect using Azo‐NZ1. Upon UV irradiation, currents recovered to 85–97%. The remaining 15–3% might be explained by not complete switching of trans‐Azo‐NZ1 into cis‐state, not complete dissociation of cis‐Azo‐NZ1 form the pore, or fast development of the receptor desensitization.

Possible mechanism of Azo‐NZ1 interaction with the channel

The 2′ residue of the TM2 domain (see Figure 5a) is a key region that determines the action of several pore blockers of Cys‐loop receptors, including cyanotriphenylborate (Rundström et al., 1994), picrotoxin (Lynch, Rajendra, Barry, & Schofield, 1995; Zhorov & Bregestovski, 2000), niflumic acid (Maleeva, Peiretti, Zhorov, & Bregestovski, 2017), and cyclothiazide (Xie, Song, Ripps, & Qian, 2008). The Azo‐NZ1 construct contains two features that are essential to impart sensitivity to the photochrome: (a) the negative charge of the sulphonate group (mimicking that of the chloride ion) and (b) the distance between this sulphonate group and the carbonyl group of the benzodiazepine core (see Figure S2). In trans‐Azo‐NZ1, this distance is optimal (~14.1 Å) to bridge the two pore‐lining hydrogen bond residues S2′ and T13′. Upon UV light irradiation, cis‐Azo‐NZ1 may either remain in the channel (assuming another binding conformation that unblocks the pore) or exit the channel (reaching the bath solution). Although with the data at hand we cannot distinguish between the two possibilities, it is feasible that cis‐Azo‐NZ1 remains partially bound in the pore, as can be suggested from Figures 3c and 7b and modelling analysis (Figure 6c).

Effect of Azo‐NZ1 on synaptic currents

Our analysis on dentate gyrus neurons from mouse brain slices demonstrated that Azo‐NZ1 decreases amplitude of GABAergic synaptic events in a light‐dependent manner. The inhibition was weaker in comparison with the one observed at heterologously expressed α1/β2/γ2 GABAA receptors. This may result from the different composition of GABA receptors receptors present in the dentate gyrus neurons, which can express up to 10 GABAA receptor subunits, as single‐cell mRNA analysis demonstrated (Brooks‐Kayal, Shumate, Jin, Rikhter, & Coulter, 1998; Coulter & Carlson, 2007). Also, synaptic and extrasynaptic GABAA receptors with distinctive properties and subunit composition can be differently modulated by the photochrome. The role of these discrepancies in the neuronal circuit modulation by Azo‐NZ1 should be clarified in future studies.

Recently, several photochromic modulators of GABAA receptors were developed (reviewed in Bregestovski, Maleeva, & Gorostiza, 2018): two potentiators (Stein et al., 2012; Yue et al., 2012) and several inhibitors (Huckvale, Mortensen, Pryde, Smart, & Baker, 2016; Lin et al., 2014, 2015; Lin, Tsai, Rajappa, & Kramer, 2018). The compounds developed by Lin et al. are required to be tethered to the receptor, which needs specific mutation of the GABA receptor, that is, genetic manipulation. The compound developed by Huckvale et al. is a soluble photochromic competitive GABA receptor antagonist based on combination of azobenzene and gabazine. Azo‐NZ1, presented in our study, exhibits a completely different mode of action, being a light‐controlled blocker of ionotropic GABA receptor channels. Moreover, Azo‐NZ1 is the first known photochromic modulator of ρ2 GABAC receptors that are highly expressed in retina. Azo‐NZ1 blocks as well α2 glycine receptors in a similar UV‐dependent manner, probably due to highly conserved amino acid sequences of pore‐forming domains of GABA and glycine receptors. Indeed, the difference in Azo‐NZ1 action on α1 and α2 glycine receptors, caused by glycine to alanine substitution in 2′ position of their TM2 domains, confirms the pore‐blocking effect of Azo‐NZ1. In contrast, Azo‐NZ1 was not able to block the 5‐HT3A receptor, a cation‐selective member of Cys‐loop receptor superfamily.

Our findings pave the way for development of a novel pharmacological toolset for photo‐modulation, as well as functional and molecular physiology studies of these receptors. Azo‐NZ1 is a freely diffusible ligand that binds to endogenous receptors, and such studies could be carried out without any genetic manipulations. Photocontrol of inhibitory neurotransmission should be readily accessible in acute preparations and in all kinds of vertebrate and invertebrate wild‐type animals using Azo‐NZ1.

CONFLICT OF INTEREST

The authors declare no conflicts of interest.

AUTHOR CONTRIBUTIONS

G.M., E.P., and A.G.‐J. performed the experiments. A.B.‐B., D.W. and K.R. synthesized the compounds. A.N.‐H. and M.A.‐P. performed the molecular modelling. F.P. and P.S. performed the mutagenesis. G.M., K.R., A.N.‐H., M.A.‐P., C.R., B.K., P.G., P.B. analysed the data and wrote the manuscript. M.A.‐P., C.R., B.K., P.G., P.B. supervised the project, reviewed, discussed, and conceptualized data.

DECLARATION OF TRANSPARENCY AND SCIENTIFIC RIGOUR

This Declaration acknowledges that this paper adheres to the principles for transparent reporting and scientific rigour of preclinical research as stated in the BJP guidelines for Design & Analysis, and Animal Experimentation, and as recommended by funding agencies, publishers and other organisations engaged with supporting research.

Figure S1.

Azo‐NZ1 interacts with glycine receptors. (A) Amino acid sequences of TM2 domains of alpha1 and 2 glycine receptor, GABA beta2 and rho2 subunits. (B) On the left a representative trace of the current mediated by alpha1 glycine receptors induced by glycine (30μM) in the absence and in the presence of Azo‐NZ1 (50μM), Vh ‐30mV; on the right – representative trace of the current mediated by alpha2 glycine receptors induced by glycine (30μM) in the absence and in the presence of Azo‐NZ1 (50μM), Vh ‐30mV. (C) Cumulative data of the Azo‐NZ1 (50μM) effect at alpha1 and alpha2 glycine receptors (Vh ‐30mV); ns – not significant, * ‐ p<0.05, ** ‐ p<0.01.

Figure S2. Azo‐NZ1 does not interact with 5HT3A receptors. (A) Representative traces recorded at application of 5‐HT (black trace), mixture of 5HT with Azo‐NZ1 (50μM) at visible light (orange) and mixture of 5HT with Azo‐NZ1 (50μM) at UV light (violet); Vh ‐30mV. (B) Cumulative data demonstrating the absence of Azo‐NZ1 effect at 5HT3A receptors when applied in concentrations 50μM (on the left, n=6) and 100μM (on the right, n=6). ns – not significant, * ‐ p<0.05.

Figure S3. Density map of the sulphonate group of Azo‐NZ1 bound in the transmembrane part of the pore of the rho2 GABACR homology model. Regions of continuous density correspond to higher sulphonate occupancy and indicate tighter sulphonate binding. Each contour line corresponds to 1.3 particles/nm3, where “particle” refers to the sulphonate group in a given binding pose, and is used to follow Azo‐NZ1 binding. Therefore, the particle density is equivalent to the number of binding poses whose sulphonate group is located in a given volume of the pore (e.g. 1.3 particles/nm3 corresponds to 16% of binding poses). (A) Trans isomer; (B) Cis isomer. A longitudinal view of the transmembrane domain is shown and the front subunit is not displayed to reveal the interior of the pore. The S2′ and T13′ positions are marked with dashed lines and the percentage of binding poses in a given position is given between parentheses. In particular, for the 2′ position discussed in the text, the number of binding poses decreases from 51% (trans) to 28% (cis), indicating that trans‐azo‐NZ1 is more likely to bind with the sulphonate at the 2′ position than the cis isomer, as inferred from the electrophysiology experiments.

Figure S4. Dimensions along the pore with trans and cis‐Azo‐NZ1. (A) Longitudinal view of the 2′‐13′ region of the pore with trans‐Azo‐NZ1. The longitudinal distance of the ligand goes from the carbonyl group of the benzodiazepine core to the sulphur atom of the sulphonate group. The pore distance is calculated between the alpha carbon atoms of T13′ and S2′. (B) Intracellular view of the 2′4 ring with the sulphonate group of trans‐Azo‐NZ1. The dashed lines between the pentagon vertexes and the oxygens represent the hydrogen bonds with S2′. (C) Extracellular view of 13′ ring with the sulphonate group of cis‐Azo‐NZ1. The dashed lines between the pentagon vertexes and the oxygens represent the hydrogen bonds with T13′. (D) Extracellular view of 9′ ring with cis‐Azo‐NZ1. The ligand distance shown spans from the carbonyl of the benzodiazepine core to the second, distal nitrogen of the azo group. The pore distances are calculated between the innermost side chain atoms of the corresponding residues.

Figure S5. Density map of the sulphonate group of trans‐Azo‐NZ1 bound in the transmembrane part of the pore of rho2 and S2′G rho2 GABACR homology models. Each contour line corresponds to 16% of binding poses. (A) Rho2 receptor; (B) S2′G rho2 receptor. Longitudinal view of the pore; the front subunit is not displayed to reveal the interior of the pore. The S2′/G2′ and T13′ regions are marked with dashed lines and the percentage of binding poses in a given position is given between parentheses. In particular, for the 2′ position discussed in the text, the number of binding poses decreases from 51% (wild‐type rho2) to 37% (S2′G mutant), suggesting that the trans isomer is more likely to bind to wild‐type rho2 than to the S2′G mutant, as inferred from the electrophysiology experiments.

Figure S6. Interactions of trans‐Azo‐NZ1 in S2′‐T13′ pore region of the GABAC receptors. (A) Longitudinal view of the pore of the S2′G rho2 mutant receptor. (B) Longitudinal view of the pore of 5 the wild‐type rho1 receptor. (C) Longitudinal view of the pore of the P2′S rho1 receptor. The porelining residues are represented as ball‐and‐sticks and carbon atoms are colored in gray. Trans‐Azo‐NZ1 is represented as sticks with carbon atoms in orange. Nitrogen and oxygen atoms are colored in blue and red, respectively. Hydrogen bonds are marked with dashed lines and hydrophobic interactions are represented as yellow surfaces.

Figure S7. Density map of the sulphonate group of Azo‐NZ1 bound in the transmembrane part of the pore of rho1 and P2′S rho1 GABACR homology models, corresponding to an open state. Each contour line corresponds to 16% of binding poses. (A) Rho1 receptor; (B) P2′S rho1 receptor. Longitudinal view of the pore where one subunit is not displayed to reveal the interior of the receptor. The P2′/S2′ and T13′ regions are marked with dashed lines and the percentage of binding poses in a given position is given between parentheses. In particular, for the 2′ position discussed in the text, the number of binding poses increases from 23% (wild‐type rho1) to 39% (P2′S mutant), suggesting a lower propensity of wild‐type rho1 to bind the trans isomer compared to the P2′S rho1 mutant.

Figure S8. α1/β2/γ2 GABAAR with trans‐Azo‐NZ1 in the transmembrane part of the pore region. (A) Density map of the sulphonate group of Azo‐NZ1. Each contour line corresponds to 16% of binding poses. The five subunits of the receptor are colour‐coded in green (α1), blue (β2) and pink (γ2); the γ2 subunit is not displayed to reveal the interior of the receptor. The trans‐azo‐NZ1 binding mode with the sulphonate bound at position 2′ appears to more likely (36%) than to 13′ (31%). (B) Longitudinal view of the pore, showing the interactions of trans‐Azo‐NZ1 in the S2′‐T13′ region. The pore‐lining residues are represented as ball‐and‐sticks and the five subunits of the receptor are colour‐coded as in (A). Trans‐Azo‐NZ1 is represented as sticks with the carbons coloured in orange, nitrogens in blue and oxygens in red. Hydrogen bonds are marked with dashed lines and hydrophobic interactions are represented as yellow surfaces. (C) Intracellular view of the pore. Only hydrogen bond interactions are shown; hydrophobic interactions are not displayed for simplicity. The subunit colour code is the same as in (A) and (B).

Figure S9. Superposition of the homology model of the GABAA receptor used in this work (Bergmann et al., 2013) (open state, solid cartoon) and two of the cryo‐EM structures in reference (Masiulis et al. 2019) corresponding to desensitized and closed states. (A) Closed state with the pore channel blocker, picrotoxinin (represented as grey transparent sticks), PDB code 6HUJ (GABAAR + GABA + Picrotoxin). (B) Open state, homology model from Bergmann et al. (C) Desensitized state, PDB code 6HUP (GABAAR + GABA + Diazepam). The five subunits of the receptor are colour‐coded in green (α1), blue (β2/ β3) and pink (γ2). The closed state structure (PDB code 6HUJ) has been solved in complex with GABA and picrotoxinin, a pore blocker like trans‐azo‐NZ1, and thus it is interesting to compare their binding determinants. Picrotoxinin extends the pore from the 2′ to the 9′ ring, whereas the longer trans‐azo‐NZ1 reaches the 13′ ring (i.e. the carbonyl group of the benzodiazepine core forms a hydrogen bond with Thr13′). The 9′ and 13′ rings are narrower in the closed than in the open state (Figure S13) and thus are compatible with binding of the smaller picrotoxinin molecule, whereas they would not be suitable to accommodate the bulkier benzodiazepine core of azo‐NZ1. Therefore, trans‐azo‐NZ1 is not expected to bind to the closed receptor. The differences in pore size between the open (Bergmann et al., 2013) and the desensitized (PDB code 6HUP) structures are not as dramatic as for the closed state and thus we cannot fully discard trans‐azo‐NZ1 binding to the desensitized state. However, desensitization is known to block chloride conduction and thus we extrapolate that it might also hinder binding of the chloride‐mimic sulphonate group of trans‐azo‐NZ1. In addition, trans‐azo‐NZ1 forms interactions in the region of the desensitization gate (i.e. a hydrogen bond with the gamma2‐Ser2′) similar to picrotoxinin. Hence, the “foot‐in‐the‐door” mechanism proposed for picrotoxinin (Gielen et al. 2015), whereby the pore blocker prevents desensitization, might also apply to trans‐azo‐NZ1.

Figure S10. Diazepam molecule showing the two possible conformers (M and P) considered in the docking calculations.

Figure S11. UV‐Vis spectra of AZO‐NZ1 (A) 50 μM in DMSO, (B) 50 μM in phosphate buffer + 0.1% DMSO from the trans isomer (black), the PSS at irradiation with UV‐light of 365 nm (purple) and the PSS at irradiation with blue light of 455 nm (blue). Cycle performance of Azo‐NZ1, (C) 50 μM in DMSO, (D) 50 μM in phosphate buffer + 0.1% DMSO). Changes in absorption at λmax of the trans isomer were measured during alternate irradiation with light of 365 nm and 455 nm. Determination of the thermal half‐lives of Azo‐NZ1 at 25 °C, (E) 50 μM in DMSO, (F) 50 μM in phosphate buffer + 0.1% DMSO.

Figure S12. α1/β2/γ2 GABAAR with Azo‐NZ1 in the classical benzodiazepine binding site. (A‐C) show the density maps of the sulphonate group of Azo‐NZ1 for the trans/M, trans/P and cis/P conformers, respectively. The density is displayed at the level of 0.021 particles/nm3. The density blobs corresponding to the poses of azo‐NZ1 inside the classical binding site are circled in orange (transconformers) or purple (cis/M), and their corresponding percentage of binding poses is indicated. The α1 and γ2 subunits of the receptor are represented in green and pink cartoon, respectively. For comparison, a molecule of diazepam and the benzodiazepine binding residues are shown, represented as sticks with the carbons coloured in gray or cyan, respectively, nitrogens in blue, oxygens in red and the C‐7 chlorine atom of diazepam in brown. (D‐F) show the most representative poses of azo‐NZ1 inside the classical benzodiazepine site for the trans/M, trans/P and cis/P conformers, respectively. Azo‐NZ1 is shown as sticks with the carbons coloured in orange (trans isomers) or purple (cis/P conformer). As in (A‐C), a molecule of diazepam and the benzodiazepine binding residues are shown for comparison, represented as sticks with the carbons coloured in gray or cyan, respectively, nitrogens in blue, oxygens in red and the C‐7 chlorine atom of diazepam in brown.

Table S1. Photochromic data of Azo‐NZ1.

Click here for additional data file.