Stoichiometry-Selective Antagonism of α4β2 Nicotinic Acetylcholine Receptors by Fluoroquinolone Antibiotics

Quinolone antibiotics disrupt bacterial DNA synthesis by interacting with DNA gyrase and topoisomerase IV. However, in addition, they have been shown to act as inhibitors of pentameric ligand-gated ion channels such as GABAA receptors and the α7 nicotinic acetylcholine receptor (nAChR). In the present study, we have examined the effects of quinolone antibiotics on the human α4β2 nAChR, an important subtype that is widely expressed in the central nervous system. A key feature of α4β2 nAChRs is their ability to coassemble into two distinct stoichiometries, (α4)2(β2)3 and (α4)3(β2)2, which results in differing affinities for acetylcholine. The effects of nine quinolone antibiotics were examined on both stoichiometries of the α4β2 receptor by two-electrode voltage-clamp recording. All compounds exhibited significant inhibition of α4β2 nAChRs. However, all of the fluoroquinolone antibiotics examined (ciprofloxacin, enoxacin, enrofloxacin, difloxacin, norfloxacin, pefloxacin, and sparfloxacin) were significantly more potent inhibitors of (α4)2(β2)3 nAChRs than of (α4)3(β2)2 nAChRs. This stoichiometry-selective effect was most pronounced with pefloxacin, which inhibited (α4)2(β2)3 nAChRs with an IC50 of 26.4 ± 3.4 μM but displayed no significant inhibition of (α4)3(β2)2 nAChRs. In contrast, two nonfluorinated quinolone antibiotics (cinoxacin and oxolinic acid) exhibited no selectivity in their inhibition of the two stoichiometries of α4β2. Computational docking studies suggest that pefloxacin interacts selectively with an allosteric transmembrane site at the β2(+)/β2(−) subunit interface, which is consistent with its selective inhibition of (α4)2(β2)3. These findings concerning the antagonist effects of fluoroquinolones provide further evidence that differences in the subunit stoichiometry of heteromeric nAChRs can result in substantial differences in pharmacological properties.


■ INTRODUCTION
Nicotinic acetylcholine receptors (nAChRs) form part of the superfamily of pentameric ligand-gated ion channels, which includes receptors for 5-hydroxytryptamine (5-HT), γ-aminobutyric acid (GABA), and glycine. 1 Seventeen nAChR subunits have been identified in vertebrates (α1−α10, β1−β4, γ, δ, and ε) that can coassemble in a variety of combinations to generate a diverse family of pharmacologically distinct nAChR subtypes, including both heteromeric subunit combinations (such as α4β2) and homomeric complexes (such as α7). 2 Further complexity can arise as a consequence of nAChR subunits coassembling with different stoichiometries. For example, the α4 and β2 subunits can coassemble into pentameric complexes containing either two α4 and three β2 subunits ((α4) 2 (β2) 3 ) or three α4 and two β2 subunits ((α4) 2 (β2) 3 ). 3 As has been reported previously, the two stoichiometries of α4β2 nAChR differ in their sensitivity to acetylcholine (ACh) and, as a consequence, are often referred to as "high-sensitivity" and "low-sensitivity" subtypes, respectively. 4 Receptors containing α4 and β2 subunits mediate the effects of nicotine associated with tobacco smoking and are the site of action of drugs used to assist with smoking cessation. 5 In addition, α4β2 nAChRs are targets for drug discovery in areas such as cognition, attention, and pain. 6−8 In recent years, considerable attention has focused on studies of allosteric modulators of nAChRs that are thought to bind within the receptor's transmembrane domain. 9,10 Quinolone antibiotics interact with two distinct targets within bacterial cells, DNA gyrase (DNAG) and topoisomerase IV, both of which are involved in bacterial DNA synthesis. 11 Quinolones inhibit DNA synthesis by stabilizing complexes of DNA and topoisomerase IV or DNAG which blocks the progression of the replication fork. 11 However, previous studies have indicated that quinolone antibiotics can also modulate pentameric neurotransmitter-gated ion channels. For example, they have been reported to inhibit ionotropic receptors for GABA (GABA A receptors) 12−15 and also human α7 nAChRs. 16 In the case of α7 nAChRs, pefloxacin was identified as a potential allosteric modulator (interacting with the α7 nAChR transmembrane domain) on the basis of virtual screening, 16 performed with a revised homology model of the α7 nAChR, 17 and was subsequently shown to act as a noncompetitive antagonist on α7 nAChRs. 16 Here, we have examined the effects of a series of nine quinolone antibiotics (Figure 1), including pefloxacin, on the two stoichiometries of the human α4β2 nAChR by two-electrode voltage-clamp recording of cloned receptor subunits expressed in Xenopus oocytes.

■ MATERIALS AND METHODS
Plasmids and Reagents. Ciprofloxacin, enrofloxacin, difloxacin, and sparfloxacin were purchased from Sigma-Aldrich (Gillingham, U.K.). Pefloxacin, cinoxacin, and oxolinic acid were purchased from Santa-Cruz Biotechnology (Dallas, TX, USA). Enoxacin was purchased from TOKU-E (Washington, USA), and norfloxacin was purchased from Merck Life Science UK Ltd. (Southampton, U.K.). Stock solutions of antibiotics (100 mM) were prepared in DMSO, with the exception of enoxacin which was prepared in 1 M NaOH. Stock solutions were stored at −20°C before use.
Plasmids and Site-Directed Mutagenesis. Human nAChR subunit cDNAs in plasmid expression vector pSP64GL (pSP64GL-α4 and pSP64GL-β2) have been described previously. 18 Site-directed mutagenesis (to generate plasmids pSP64GL-α4 L283A , pSP64GL-α4 S284A , and pSP64GL-β2 V278A ) was performed using the QuikChange mutagenesis kit (Agilent Technologies) and verified by nucleotide sequencing (Source Bioscience). Note that the numbering of these amino acids in the human nAChR α4 and β2 subunits is based on the intact protein sequence (including the signal sequence), as indicated in the EMBL/GenBank database entries NP_000735.1 and NP_000739.1, respectively.
RNA Synthesis and Oocyte Expression. Plasmid expression vectors were linearized by restriction enzyme digestion at sites downstream from the inserted cDNA. Linearized plasmids were purified with QIAQuik PCR purification kit (Qiagen), and transcription of cRNA was carried out using mMESSAGE mMACHINE Figure 1. Chemical structures of quinolone antibiotics. The effects of seven fluoroquinolone antibiotics (ciprofloxacin, difloxacin, enoxacin, enrofloxacin, norfloxacin, pefloxacin, and sparfloxacin) and two nonfluorinated quinolone antibiotics (cinoxacin and oxolinic acid) were examined in the present study.
Oocytes were injected, using a Drummond variable volume microinjector, with 32.2 nL of cRNA containing either a mixture of 30 ng/μL human α4 and 300 ng/μL human β2 or 300 ng/μL human α4 and 30 ng/μL human β2 cRNAs.
Oocyte Electrophysiology. Adult female Xenopus laevis frogs were obtained from the European Xenopus Resource Centre at the University of Portsmouth. Animals were sacrificed using Schedule 1 Figure 2. Inhibitory effects of quinolone antibiotics on α4β2 nAChRs: bar charts illustrating the effects of quinolone antibiotics on (α4) 2 (β2) 3 nAChRs (white bars) and (α4) 3 (β2) 2 nAChRs (black bars) expressed in Xenopus oocytes. Antibiotics (100 μM) were preapplied for 30 s and then coapplied with agonist (an EC 50 concentration of ACh) for 5 s or until a plateau in the response. Responses are normalized to responses to ACh in the absence of antibiotic. Data are the mean ± SEM from at least three individual experiments (as indicated). Significant differences are indicated ( * * = P < 0.01, * * * = P < 0.001, ns = not significant).
procedures approved by the Animals (Scientific Procedures) Act 1986 and by the UCL Research Ethics Committee. Xenopus were anesthetized by immersion in 0.2% MS222 for 15 min (or until complete anesthesia was confirmed by absence of leg-withdrawal and righting reflex), followed by cranial concussion, decapitation, and pithing. Xenopus oocytes were isolated, maintained, and injected with cRNA, as described previously. 21 Two-electrode voltage-clamp recordings were performed using a Warner Instruments OC-725C amplifier (Harvard Apparatus) with the oocyte membrane potential held at −60 mV, as described previously. 22 Oocytes were continuously perfused with a modified Ringer's solution (115 mM NaCl, 2.5 mM KCl, 1.8 mM BaCl 2 , and 10 mM HEPES, pH 7.3). Application of compounds was controlled by LabChart software (AD Instruments) using a BPS-8 solenoid valve solution exchange system (ALA Scientific Inc.). Typically, agonists were applied for 5 s or until a plateau in the response was observed. Antagonists were preapplied for 30 s and then coapplied with agonist for 5 s or until a plateau in the response was observed. Where data has been normalized to a maximum ACh response, the maximum response was determined from a minimum of three independent ACh dose−response curves.
Statistical Analysis. For individual pairwise comparisons, statistical significance was determined using unpaired Student's t tests or ANOVA for multiple comparisons. Dose−response curves were fitted by GraphPad Prism, using the following equation (where I is the current, I max is the maximum current, the EC 50 is the concentration of agonist that elicits a half-maximal response, and n H is the Hill coefficient): Small Molecule Docking. To identify potential binding sites for quinolone antibiotics in the human (α4) 2 (β2) 3 and (α4) 3 (β2) 2 nAChRs, computational docking was performed with protein structures that have been determined previously by cryoelectron microscopy (Protein Data Bank codes 6CNJ and 6CNK, respectively). 23 Small molecule computer docking was performed using AutoDock Vina (Molecular Graphics Lab at Scripps Research Institute, La Jolla, CA) and PLANTS (Protein−Ligand ANT System; Universitaẗ Tubingen, Germany). Docking was performed within a search area of 18 Å radius centered on the γ-carbon of T286 (α4) or T277 (β2) of the subunit corresponding to the principal (+) side of the subunit interface. This covered the inter-and intrasubunit cavities of the β2/α4 and α4/α4 interfaces of (α4) 3 (β2) 2 (PDB code 6CNK) and the β2/α4 and β2/β2 interfaces of (α4) 2 (β2) 3 (PDB code 6CNJ). With both docking programs, ligands were allowed to be fully flexible and the maximum search efficiency was used. One-thousand protein−ligand conformations were produced by each docking program for each interface query and analyzed with a previously described consensus docking protocol. 17 This in-house script allows for a consensus binding mode or cluster to be identified from the protein−ligand conformations produced from the two independent docking programs. The rationale for this approach is to identify predicted binding sites for which there is a consensus between two docking programs that employ different scoring functions. The most highly populated consensus cluster of solutions (determined by RMSD with a cutoff of 2 Å between the two docking programmes) and highest ranked (by either PLANTS or AutoDock Vina scoring function) was taken to represent the active conformation of the ligand in each receptor stoichiometry.
Docking of Quinolone Antibiotics into α4β2 nAChR Structures. Computational docking studies were performed with three-dimensional atomic models of the (α4) 2 (β2) 3 and (α4) 3 (β2) 2 nAChRs that had been determined previously by cryoelectron microscopy (PDB codes 6CNJ and 6CNK, respectively). 23 A consensus docking approach 17 was employed, involving two independent docking methods (Auto-Dock Vina and PLANTS). Since previous studies had identified the intersubunit transmembrane region as being the most plausible binding site for allosteric modulators such as pefloxacin in the α7 nAChR, 16,17 docking studies were performed within a search area of 18 Å radius centered in this region (see Materials and Methods). When results were compared from the two computational docking studies, no consensus binding site for pefloxacin was identified in the (α4) 3 (β2) 2 nAChR subtype, whereas a single plausible consensus binding site was identified in (α4) 2 (β2) 3 at the β2/β2 interface ( Figure 5) at a location similar to that identified previously for allosteric modulators of nAChRs. 16,17 These findings are consistent with evidence that pefloxacin is a selective antagonist of the (α4) 2 (β2) 3 nAChR subtype. In contrast, docking studies with cinoxacin identified plausible binding sites in both receptor structures. Again, this is consistent with the finding that these compounds display no selectivity in their antagonist effects on (α4) 2 (β2) 3 and (α4) 3 (β2) 2 nAChRs. Three binding sites were identified within (α4) 2 (β2) 3 (one within the β2/β2 interface and two within the β2/α4 interface) ( Figure 5), and two binding sites were identified in the (α4) 3 (β2) 2 nAChR (both within the β2/ α4 interface) ( Figure 5).
Further docking studies were performed with the other seven quinolone antibiotics that had been examined on nAChRs expressed in Xenopus oocytes (ciprofloxacin, difloxacin, enoxacin, enrofloxacin, norfloxacin, oxolinic acid, and sparfloxacin). These are compounds that, like cinoxacin, displayed antagonist effects on both α4β2 stoichiometries. As was observed with docking studies with cinoxacin (but in contrast to pefloxacin), plausible binding sites were identified for all seven of these compounds in both α4β2 stoichiometries and in positions that closely resembled those that had been identified with cinoxacin.
Effects of Pefloxacin and Cinoxacin on Mutant α4β2 nAChRs. A possible explanation for the nonselective antagonism by compounds such as cinoxacin and for the selective antagonism by pefloxacin might be that cinoxacin is able to bind to subunit interfaces containing the α4 subunit, whereas pefloxacin binds selectively at the interface of two β2 subunits. Such an explanation would also be consistent with the computer docking studies. With the aim of testing this hypothesis, the influence of α4 subunit mutations was examined on the antagonist effects of pefloxacin and cinoxacin. Two amino acids within the transmembrane domain of the α4 subunit were selected for site-directed mutagenesis (L283 and S284) due to their close proximity to the predicted binding sites of cinoxacin and the lack of proximity to the predicted binding site for pefloxacin. A further reason for selecting these two amino acids was that mutagenesis of the analogous amino acids in α7 nAChRs has been shown to alter allosteric modulation by compounds such as pefloxacin. 16 Both amino acids were mutated individually to alanine to create α4 L283A and α4 S284A . In addition, an amino acid within the transmembrane domain of the β2 subunit (V278) was selected for site-directed mutagenesis due to its proximity to the predicted binding site of both cinoxacin and pefloxacin and was mutated to alanine to create β2 V278A .

■ DISCUSSION
A notable aspect of the present study is that fluoroquinolone antibiotics exhibit stoichiometry-selective antagonism of α4β2 nAChRs. The effect was most pronounced for pefloxacin, which exhibits complete selectivity for α4β2 nAChRs in the (α4) 2 (β2) 3 stoichiometry. The primary difference between the two α4β2 nAChR stoichiometries is the presence of a β2/β2 interface in (α4) 2 (β2) 3 and an α4/α4 interface in the (α4) 3 (β2) 2 . It is of interest, therefore, that computational docking studies are consistent with the possibility that pefloxacin binds preferentially to a site at the β2/β2 interface in (α4) 2 (β2) 3 nAChRs, whereas less selective and nonselective quinolone antibiotics were predicted to interact with sites at both the β2/β2 and β2/α4 subunit interfaces. Although plausible binding sites were identified in (α4) 2 (β2) 3 for all nine quinolone antibiotics examined, the predicted binding site for pefloxacin is qualitatively distinct from that of the other compounds, extending deeper into the intersubunit cavity within the β2/β2 subunit interface. In addition, while cinoxacin and the other quinolone antibiotics are predicted to interact with TM2 of both subunits at the β2/β2 interface, pefloxacin is predicted to also interact with the TM1 and TM3 helices of the complementary and primary subunits, respectively. This supports the possibility that pefloxacin may make important interactions with the β2/β2 interface that are distinct from that of the other antibiotics examined.
A prediction, based on our docking results, was that mutations in the α4 subunit and close to the predicted intersubunit transmembrane binding site of quinolone antibiotics might have a more profound effect on nonselective antibiotics such as cinoxacin (that were predicted to bind at both the β2/α4 and β2/β2 subunit interfaces) than pefloxacin (that was predicted to bind exclusively at the β2/β2 subunit interface), having found that two such mutations (α4 L283A and α4 S284A ) abolish the antagonist effects of cinoxacin but have no significant effect on pefloxacin supports the predictions. These particular amino acids were selected for mutagenesis studies because they are at positions in the α4 subunit that are analogous to two amino acids in the α7 nAChR (α4 S248 and α7 L247 ) that have been shown previously to modulate the effects of compounds predicted to bind in the intersubunit transmembrane cavity. 16 In addition, a mutation was made within the β2 subunit (V278A) at a site that is in close proximity to the predicted binding site of both cinoxacin and pefloxacin. As was seen with nAChRs containing α4 transmembrane mutations, the inhibitory effects of cinoxacin were abolished by this mutation in both stoichiometries. Interestingly, the β2 V278A mutation converted pefloxacin but not cinoxacin into a partial agonist. There are previous examples of transmembrane mutations converting antagonists into agonist, a finding that is probably a consequence of the mutations causing conformational changes that alter the energy barrier for transitions between open and closed states following ligand binding or by allowing bound ligands to more easily stabilize the open conformation. One of the best characterized examples is a transmembrane mutation in the nAChR α7 subunit (L247T) that causes increase spontaneous openings, reduces receptor desensitization, alters temperature sensitivity, and converts antagonists into agonists. 24−27 Similarly, this α7 nAChR mutation can convert both positive allosteric All data are normalized to responses to an EC 50 concentration of ACh and are the mean ± SEM of at least three independent experiments. Significant differences are indicated ( * * * = P < 0.001, ns = not significant). modulators and silent allosteric modulators into allosteric agonists. 28−30 In addition, several other nAChR transmembrane mutations have been reported that convert positive allosteric modulators into either agonists or antagonists. 17,31 It is of interest that whereas some degree of selectivity for (α4) 3 (β2) 2 nAChRs was observed with all fluoroquinolone antibiotics, no selectivity between (α4) 2 (β2) 3 and (α4) 3 (β2) 2 was seen with the two nonfluorinated quinolone antibiotics (cinoxacin and oxolinic acid). However, in contrast to the situation with pefloxacin, our docking studies do not provide a comprehensive explanation for this difference. Indeed, all of these antibiotics (with the exception of pefloxacin) were predicted to bind in broadly similar locations. However, it may be worth noting that cinoxacin and oxolinic acid were predicted to bind at positions in the transmembrane intersubunit cavity, while larger fluoroquinolones bound at position closer to the central ion channel and extended into the pore. Our findings extend previous evidence demonstrating that a variety of nicotinic ligands can show selectivity for the different stoichiometries of α4β2 nAChRs. This includes evidence for the stoichiometry-selective modulation of α4β2 nAChRs by agonists, 3,4,19,20,32−34 competitive antagonists, 3,19 divalent cations, 32,35 and positive allosteric modulators. 36−42 It has been estimated that between 1% and 4% of individuals treated with quinolone antibiotics display adverse side effects, including headaches, insomnia, and in some cases convulsions that become more prevalent when quinolone antibiotics are coadministered with nonsteroidal anti-inflammatory drugs. 43−45 It has been suggested that these side effects are mediated via interactions with GABA A receptors, since inhibitors of these receptors are proconvulsant, whereas potentiators are anxiolytic and sedative. 46 Radioligand binding experiments have demonstrated that quinolone antibiotics can inhibit the binding of [ 3 H]GABA or [ 3 H]muscimol to GABA A receptors in preparations of rat or mouse brain synaptic membranes. Furthermore, this inhibition was shown to be more potent when the antibiotics were coadministered with biphenylacetic acid, a nonsteroidal anti-inflammatory drug. 12,13 Subsequently, whole-cell voltage-clamp recordings of rat dorsal root ganglion neurons and hippocampal neurons have demonstrated inhibition of GABA-evoked responses of GABA A receptors by quinolone antibiotics, an effect that was also increased by the presence of biphenylacetic acid. 13,15,47 In contrast, radioligand binding experiments have shown no effects of quinolone antibiotics on agonist binding to excitatory glutamate receptors, muscarinic acetylcholine receptors, and GABA B receptors. 48,49 It is unclear whether the antagonist effects of quinolone antibiotics observed on nAChRs have any relevance to the side effects that are sometimes reported, but it is of interest that they can exert significant effects on both inhibitory GABA A receptors and excitatory nAChRs, both members of the superfamily of pentameric ligand-gated ion channels.
In previous studies, pefloxacin has been shown to be a noncompetitive antagonist of α7 nAChR and was originally identified on the basis of virtual screening for compounds predicted to interact with an allosteric transmembrane site on the α7 nAChR. 16 Here we have obtained evidence of insurmountable antagonism with both pefloxacin and cinoxacin on α4β2 nAChRs that is consistent with them acting as noncompetitive antagonists of α4β2 nAChRs. It is well-known that the pharmacological properties of nAChRs are influenced by subunit composition, but the present study provides further evidence that such properties can also be influenced by the same subunits being arranged in different stoichiometries.