The role of loops B and C in determining the potentiation of GABAA receptors by midazolam

Abstract Many benzodiazepines are positive allosteric modulators (PAMs) of GABAA receptors that cause sedation, hypnosis, and anxiolysis. Benzodiazepines bind GABAA receptors at the extracellular interface of the α and γ subunits. Within the α subunit, the benzodiazepine binding site is defined by three highly conserved structural loops, loops A‐C. Although previous mutagenesis studies have identified His102 in Loop A as important for benzodiazepine modulation of GABAA receptors, the functional roles of many of the other conserved residues in loops A‐C remain incompletely understood. In this study, we made single mutations in loops A‐C of the benzodiazepine binding‐site across all six α subunits. We used whole‐cell patch clamp recording to measure the functional effects of these mutations on midazolam potentiation. The results showed that mutating the threonine in loop B and serine in loop C (Thr163 and S206 in human α1) did not abolish the receptors’ responsiveness to midazolam, as the α1(H102R) mutation did. The loop C mutations exhibited a novel array of α‐isoform specific effects on midazolam potentiation. The α3(S230I) and α5(S209I) mutations had the largest effect on midazolam potentiation, increasing the efficacy of midazolam. Novel benzodiazepines targeting loop C may represent a future direction for designing new drugs that specifically alter the activity of α3‐ and α5‐containing GABAA receptors.

membrane potentials in the adult mammalian brain. One highly relevant class of positive allosteric modulators (PAMs) of GABA A receptors are benzodiazepines.
Benzodiazepines bind at the extracellular interface of the α and γ subunits. 6 There are three structural loops (loops A, B, and C) on the α-subunit and three loops on the γ2 subunit (loops D, E, and F) that form the structure of the benzodiazepine binding site ( Figure 1B).
Loops A-C form connectors between sequential β-strands. They are sometimes referred to as loop 5 (loop A), loop 8 (loop B) and β-sheet 10 (loop C), based on nomenclature for the acetylcholine-binding protein. 7,8 Loops A-C are highly conserved across GABA A receptor subunits and form a homologous GABA agonist binding site at the β+/αinterface. 9,10 A combination of mutagenesis with functional or binding assays has been used to determine the role of specific amino acids within the structural loops A-F of the benzodiazepine site. [11][12][13][14][15][16] The conserved histidine in loop A (His101 in rodents and His102 in bovine and human cDNAs) is important for the molecular and behavioral actions of diazepam using in vitro experiments 6,11 and knock-in mice. 2,17 Other residues in loops A-C have been studied, but most mutagenesis experiments were constrained to mutating less than three α subunit isoforms. This limits the conclusions drawn. Many benzodiazepine ligands bind to multiple GABA A receptor assemblies, and a mutagenesis study across the six α subunits is needed to determine the structural role of specific residues on benzodiazepine efficacy and potency.
In this study, we examined two residues within the conserved loops B and C across all six α subunits. The conserved threonine in loop B (GSYAYTR) and serine in loop C (SSTGEYV) have been reported to differentially affect the potency and efficacy of benzodiazepine-site ligands, including that of zolpidem, eszopiclone, flumazenil, and β-carbolines. 12,[18][19][20][21] It is less understood how these specific residues affect the functional actions of nonspecific positive benzodiazepines across the six human α subunits. In this study, we mutated the highly conserved histidine in loop A (His102 in α1), threonine in loop B (Thr163 in α1), and serine in loop C (Ser206 in α1) in all six GABA A α subunits. The α4 and α6 subunits have different residues (R100, P161, and I/N204) in these locations ( Figure 1C) and form GABA A receptors insensitive to classic benzodiazepines, historically known as diazepam-insensitive receptors. 22 If midazolam acts as a canonical benzodiazepine then canonical mutations in α1-3 and α5 to residues present in α4 and α6 should block its actions and vice versa in α4 and α6. Whole-cell patch clamp recording was used to measure the actions of midazolam on mutated α x β 2 γ 2s GABA A receptors.
Midazolam was selected for this study because it is commonly used in the clinic to induce sedation, 23 it is easier to handle than other benzodiazepines (lower affinity for diazepam-sensitive receptors and higher solubility), and knowledge of its pharmacology could provide insight into designing novel sedatives with fewer side effects. We found that mutating the threonine and serine in loop B and loop C altered the efficacy of midazolam less than mutating the histidine in loop A across α1-6. Surprisingly, mutating the serine in loop C altered F I G U R E 1 The structural loops A-C within the α subunit form the benzodiazepine binding site on the GABA A receptor. (A) The assembly of the α x β 2 γ 2 GABA A receptor with arrows pointing to the two GABA sites (black) and highaffinity benzodiazepine site (red). (B) The structural loops A-C (blue, magenta, cyan) on the α subunit and loops D-F (grey) on the γ subunit form the benzodiazepine site (red dotted circle) on the α x β 2 γ 2 receptor. Target residues used in this study noted under loops. (C) The structural loops A-C are highly conserved across GABA A receptor α subunits. The location of the residues of interest are highlighted in bold with the specific mutation numbers listed to the right. The numbering is based on the human mature peptide sequences not including the signal peptide (peptide sequences based on NP_000797 (α1), NP_000798 (α2), NP_000799 (α3), NP_000800 (α4), NP_000801 (α5), NP_000802 (α6)). The mutations made in this study are referred to by the abbreviations "loop A", "loop B" and "loop C" in subsequent figures and text the efficacy of midazolam potentiation in different directions depending on the α isoform. These subunit-selective observations will be useful for the design of α3and α5-selective benzodiazepines.

| cDNA plasmids and mutagenesis
Human (Homo sapiens) GABA A subunits (α1-6, β2, γ2s) were subcloned into pcDNA3.1+ vectors with a cytomegalovirus (CMV) promoter. The hβ2 and hα3 sequences were humanized rat (Rattus norvegicus) cDNA with amino acid substitutions made to match the human protein sequence. The α1-3, α5, β2, and γ2s subunits were a generous gift from Neil L. Harrison (Columbia University Medical Center, NY). The α4 subunit was obtained from GenScript (Piscataway, NJ), and the α6 subunit was a generous gift from Robert L.
McDonald (Vanderbilt University, TN). All point mutations (listed in Figure 1C) were introduced using the QuikChange Lightening sitedirected mutagenesis kit (Agilent Technologies, Santa Clara, CA) according to the manufacturer's instructions and were confirmed by sequencing (Eurofins MWG Operon, Louisville, KY).

| In vitro electrophysiology
Wildtype and mutant GABA A receptors were characterized using whole-cell voltage-clamp electrophysiology of HEK293T cells expressing α x β 2 γ 2s receptors and GFP, similar to methods previously described. 25 Patch pipettes were created from thin-walled borosilicate glass (TW150F-4, World Precision Instruments, Inc., Sarasota, FL) using a horizontal puller (P97, Sutter Instruments, Inc., Novato, CA) to give a resistance of 2-8 MΩ when filled with intracellular solution (120 mM KCl, 2 mM MgCl 2 , 10 mM EGTA, 10 mM HEPES, and adjusted to pH 7.2 with NaOH, 315 mOsm). Extracellular solution contained 161 mM NaCl, 3 mM KCl, 1 mM MgCl 2 , 1.5 mM CaCl 2 , 10 mM HEPES, and 6 mM D-glucose, adjusted to pH 7.4 with NaOH (320-330 mOsm). GABA and midazolam (Hospira, Lake Forest, IL) were delivered using a rapid solution changer (RSC-160, BioLogics Science Instruments, Seyssinet-Pariset, France) connected to a 10-channel infusion pump (KD Scientific Inc., Holliston, MA).  Figure S1 for waveform of drug exposure). GABA preand postcontrol runs were performed before and after each midazolam assay for each cell to verify a consistent EC 10 GABA response and full washout of midazolam. Control runs consisted of 3 seconds of GABA (EC 10 ) and then 3 seconds of a saturating GABA concentration (100-300 μM depending on the α subunit) with 8 seconds of washout between ligand applications. Cells were recorded with the midazolam protocol no more than two times to avoid desensitization and incomplete washout or irreversible modulation.

| Statistics
Optimal sample sizes (n ≥ 10 cells) were calculated beforehand from preliminary α1 mutant data using G*Power (Heinrich-Heine-Universität Düsseldor, Germany) (α = 0.05 and β = 0.8) for a one-way analysis of variance (ANOVA) test. Hill parameters (maximum response or potentiation, Hill coefficient, EC 50 ) from concentration-response curves (GABA and midazolam each) were compared for significant differences within each α subunit (α1-6) and its loops A-C mutants using a one-way ANOVA at the significance threshold of α = 0.05.
Where the results of the ANOVA were significant (P < 0.05), Dunnett's post-hoc analysis for multiple comparisons (α = 0.05) was performed. Statistical analysis was carried out using Prism 7.0 (Graphpad Software, Inc., La Jolla, CA).

| RESULTS
We hypothesized that mutating single residues in the conserved loops A-C of the benzodiazepine binding site ( Figure 1C) would alter the modulation of GABA A receptors by midazolam. Whole-cell patch clamp recording of α1-6-containing α x β 2 γ 2s GABA A receptors was used to measure the degree of potentiation by midazolam within the therapeutically relevant range of 10-1000 nM. [26][27][28] Midazolam potentiation was measured as the percent of enhancement in GABA-evoked currents. A 100% potentiation was a doubling in amplitude of the whole-cell current relative to the control EC 10 GABA-response. Interestingly, we found that loop C mutations in α3 and α5 GABA A subunits increased the maximum potentiation by midazolam. However, single residue mutations in loop B and loop C did not alter, abolish or confer midazolam sensitivity as dramatically as the histidine-to-arginine exchanges in loop A.

| Loop A mutations
The loop A mutation substituted the highly conserved histidine residue for an arginine residue (FFHNG) in the α1, α2, α3, and α5 subunits. For the α4 and α6 subunits, the reverse arginine-to-histidine mutation was made. GABA concentration-response assays revealed only modest changes in loop A mutant receptors ( Figure 2, Table S1). The presence of the arginine right-shifted the GABA con-  (Table 1, see Table S2 for midazolam potentiation values). This is consistent with previous reports using diazepam. 11 The α4(R100H) and α6(R100H) mutations conferred the ability to receptors to respond to midazolam potentiation (midazolam EC 50 : α 4 (R100H)β 2 γ 2 = 73.99 ± 3.44 nM (n = 8) and α 6 (R100H)β 2 γ 2 = 41.88 ± 6.02 nM (n = 7), Figure 3C). The wildtype α 4 β 2 γ 2 and α 6 β 2 γ 2 receptors showed no notable midazolam potentiation, and no meaningful Hill parameters could be estimated (Figure 3A-B, see Table S2 for values). Confirming the role of this histidine in loop A with midazolam provided a reference for how altering a key structural residue in a conserved region of the benzodiazepine binding site can maximally alter the amplitude of midazolam potentiation of the α x β 2 γ 2 GABA A receptors.

F I G U R E 3 Mutations in loops
Overall, mutating the threonine (loop B) and serine (loop C) residues failed to dramatically abolish the ability of α x β 2 γ 2 GABA A receptors to be modulated by midazolam, as has been established for the critical histidine in loop A. Mutations in loop C had a novel array of effects on midazolam efficacy, particularly for α3and α5-containing GABA A receptors.

| DISCUSSION
Midazolam is a benzodiazepine used to induce sedation and anesthesia. 23 The therapeutically relevant range of midazolam measured from plasma is 75 ng/mL (207 nM, postoperative drowsiness) to 350 ng/mL (966 nM, anesthetized state). 26-28 PAM benzodiazepines were initially thought to enhance the activity of GABA A receptors by altering the GABA binding steps, 29 but more recent models have focused on gating mechanisms. 30,31 The structural loops, loops A-C within the α subunit, define half of the benzodiazepine site on GABA A receptors. Understanding how different parts of the benzodiazepine site interact with modulators will help us better define the precise molecular mechanisms of these drugs.
In this study, we examined the role of the histidine in loop A, thre-  Example trace for α 2 β 2 γ 2 and α 2 (S205I)β 2 γ 2 receptors. Scale bar is 5 seconds, 500 pA. (B) Example trace for α 3 β 2 γ 2 and α 3 (S230I)β 2 γ 2 receptors. Scale bar is 5 seconds, 320 pA for α 3 β 2 γ 2 and 5 seconds, 500 pA for α 3 (S230I)β 2 γ 2 receptors. The dotted line marks the highest degree of midazolam potentiation for each example trace. (C) Quantifying the amplitude of maximum potentiation in the presence of 1 μM midazolam for α 2 β 2 γ 2 , α 2 (S205I)β 2 γ 2 , α 3 β 2 γ 2 and α 3 (S230I)β 2 γ 2 receptors. *P < 0.05 significance was determined using a two-way ANOVA with Sidak's post hoc analysis. Bars are mean ± SEM from n = 7-8 cells per group Across the 18 mutations made in loops A-C within the benzodiazepine site, only subtle changes were seen in GABA apparent-affinity. Since the mutation was away from the GABA binding site, it is unlikely the mutations caused a structural rearrangement of the extracellular domain that affected the channel's activation. The α6 (N204I) mutant increased the GABA's apparent-affinity, but this was not sufficient to make the receptor any more responsive to midazolam than the wildtype α6-containing receptors. On the whole, our results were consistent with mutations that had minimal effects on GABA's normal actions at the mutated receptor.
It is well-established that the conserved histidine present in loop A (FFHNG) of the α subunit is important in determining the molecular 6,11,33 and behavioral 1 effects of benzodiazepines. This histidine is present in the α subunits sensitive to positive benzodiazepines, but in α4 and α6 isoforms that are insensitive, an arginine is present that sterically inhibits benzodiazepines from interacting properly with the receptor. 15,22 In our study, the histidine-to-arginine mutations in The loop C mutation had more obvious changes in the efficacy of midazolam potentiation. The wildtype α1, α2, α3, and α5 subunits all contain the homologous Ser206 (human α1) that we predicted would reduce midazolam potentiation when mutated to an isoleucine. Surprisingly, the results did not follow the predicted pattern. In the α1(S206I) and α2(S205I) mutants, the isoleucine decreased midazolam's maximum potentiation by 31-33%, but in α3(S230I) and α5 (S209I), it increased midazolam's potentiation by approximately 63%.
Only α3(S230I) significantly (P < 0.05) altered midazolam's EC 50. In the case of an allosteric modulator, an altered EC 50 might be caused by changes in the modulator's ability to bind and interact with the receptor or the modulator's ability to alter GABA's binding and gating of the channel. 34 As mentioned above, only modest changes in GABA apparent-affinity were seen for loop C mutations, suggesting that any changes in midazolam potentiation were more likely caused by an altered midazolam-receptor interaction and not global alterations in structure that transmitted to the GABA binding site.
Loop C is important for ligand binding because it has more mobility than the other loops 35 and may affect benzodiazepine ligand selectivity. 36 Previous studies found that the α6(Asn204) and α4(Ile203) residues (both homologous to human α1(Ser206)) were important for distinguishing the binding of negative benzodiazepines. 19 Ser206 also physically interacts with diazepam in α1, α2 and α5, suggesting a critical role in benzodiazepine action. 37 However, a neighboring mutation, homologous to α1(T207C), specifically altered benzodiazepine efficacy and not binding. 12 We propose that the homologous Ser206 in loop C may provide an important point of contact between the ligand and benzodiazepine site that affects the coupling of the benzodiazepine site to GABA activation, thereby affecting the benzodiazepine's efficacy. Because the effect of mutations in α3 and α5 were most dramatic, this serine may be more appropriately positioned in these subunits to alter midazolam's efficacy.
The α3 and α5 subunits have specific expression profiles in the brain that reflect their roles in cognitive-and limbic-related pathways. The α3 subunit is expressed in the cortex, amygdala, olfactory bulb, and thalamic reticular nucleus, where α 3 β 2/3 γ 2 receptors mediate phasic inhibition. The α5 subunit is most highly expressed in the pyramidal hippocampal cells but also in the cortex and hypothalamus. 4,38 The α 5 β 3 γ 2 receptors contribute to tonic inhibition in the hippocampus 39 and have increasingly been studied for their role in cognition 40,41 and anesthetic-induced neurotoxicity. 42 In our results, the greatest increase in midazolam's efficacy was seen with the α3(S230I) loop C mutation. The wildtype α3-containing receptors were the most sensitive to modulation by midazolam with the lowest midazolam EC 50 and highest maximum potentiation relative to the other α subunits. This is consistent with previous studies where diazepam and flunitrazepam potentiated α 3 β 1 γ 2 receptors more than α 1 β 1 γ 2 receptors. 43,44 Even with the higher wildtype levels of midazolam potentiation, the α3(S230I) loop C results were still notable. The α3(S230I) mutation in loop C dramatically increased the efficacy of midazolam potentiation compared to α2(S205I) (Figure 4) despite both α 2 β 3 γ 2 and α 3 β 3 γ 2 wildtype receptors having similar GABA apparent-affinities ( Figure S2). This novel finding underlines the importance of better understanding the differences in allosteric modulation of GABA A receptors expressing α3 compared to other α subunits. For example, nonhypnotic drugs targeting the α2 and α3 subunits have been studied for their anxiolytic and analgesic effects. 41,45 However, creating ligands that distinguish these two subunits remains difficult, as shown when an "α3-specific" PAM (SB-205384) was found to potentiate α6-containing GABA A receptors even more strongly than α3. 46 Another way to distinguish different GABA A receptor subtypes is through the γ subunit. Although other γ subunits can form benzodiazepinesensitive receptors, the γ3 subunit is less prevalent (~14% of receptors), 47 and the γ1 subunit notably reduces the benzodiazepine affinity of the receptor. 48 The γ2 subunit is the major γ isoform expressed in native GABA A receptors, 49 and thus α x β 2 γ 2 receptors provide a reasonable estimate of benzodiazepine efficacy in the brain. Based on our results, loop C might be a potential target for developing novel drugs that specifically modulate α3and α5-containing GABA A receptors using PAMs targeting the allosteric benzodiazepine site.