Engineered Flumazenil Recognition Site Provides ... - ACS Publications

Subscriber access provided by University of Sussex Library

Article

Engineered flumazenil recognition site provides mechanistic insight governing benzodiazepine modulation in GABA receptors A

David C. B. Siebert, Konstantina Bampali, Roshan Puthenkalam, Zdravko Varagic, Isabella SartoJackson, Petra Scholze, Werner Sieghart, Marko D. Mihovilovic, Michael Schnürch, and Margot Ernst ACS Chem. Biol., Just Accepted Manuscript • DOI: 10.1021/acschembio.8b00145 • Publication Date (Web): 10 Jul 2018 Downloaded from http://pubs.acs.org on July 12, 2018

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 27 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

ACS Chemical Biology

Engineered flumazenil recognition site provides mechanistic insight governing benzodiazepine modulation in GABAA receptors

David C. B. Siebert (1)§, Konstantina Bampali (2)§, Roshan Puthenkalam (2), Zdravko Varagic (2), Isabella Sarto-Jackson (3), Petra Scholze (4), Werner Sieghart (2), Marko D. Mihovilovic (1), Michael Schnürch (1), Margot Ernst (2)*

(1) Institute of Applied Synthetic Chemistry, TU Wien, Getreidemarkt 9/163, 1060 Vienna, Austria (2) Department of Molecular Neurosciences, Center for Brain Research, Medical University Vienna, Spitalgasse 4, 1090 Vienna, Austria. (3) Konrad Lorenz Institute for Evolution and Cognition Research, Altenberg, Austria (4) Department of Pathobiology of the Nervous System, Center for Brain Research, Medical University of Vienna, Spitalgasse 4, 1090 Vienna, Austria

Contact: [email protected]

§

These authors contributed equally to this work.

1 ACS Paragon Plus Environment

ACS Chemical Biology 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 27

Table of contents graphic

Abstract The

anxiolytic,

anticonvulsant,

muscle-relaxant

and

sedative-hypnotic

effects

of

benzodiazepine site ligands are mainly elicited by allosteric modulation of GABAA receptors via their extracellular αx+/γ2− (x = 1,2,3,5) interfaces. In addition, a low affinity binding site at the homologous α+/β− interfaces was reported for some benzodiazepine site ligands. Classical benzodiazepines and pyrazoloquinolinones have been used as molecular probes to develop structure activity relationship models for benzodiazepine site activity. Considering all possible α+/ β− and α+/ γ− interfaces, such ligands potentially interact with as many as 36 interfaces, giving rise to undesired side effects. Understanding the binding modes at their binding sites will enable rational strategies to design ligands with desired selectivity profiles. Here, we compared benzodiazepine site ligand interactions in the high affinity α1+/γ2− site with the homologous α1+/β3− site using a successive mutational approach. We incorporated key amino acids known to contribute to high affinity benzodiazepine binding of the γ2− subunit into the β3− subunit, resulting in a quadruple mutant β3(4mut) with high affinity flumazenil (Ro 15-1788) binding properties. Intriguingly, some benzodiazepine site ligands displayed positive allosteric modulation in the tested recombinant α1β3(4mut) constructs while diazepam remained inactive. Consequently, we performed in silico molecular docking in the wildtype receptor and the quadruple mutant. The results led to the conclusion that different benzodiazepine site ligands seem to use distinct binding modes, rather than a common binding mode. These findings provide structural hypotheses for the future optimization of both benzodiazepine site ligands, and ligands that interact with the homologous α+/β− sites. 2 ACS Paragon Plus Environment

Page 3 of 27 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


GABAA receptors are major transmitter receptors of the cys-loop receptor superfamily.1, 2 The family of GABA-gated anion channels is generally composed of five identical or different subunits. In mammals 19 different genes encode distinct subunits and therefore, a high number of GABAA receptor subtypes can be formed.2 The majority of GABAA receptors are thought to consist of one gamma (γ), two alpha (α), and two beta (β) subunits.2 In this prototypical receptor there are two extracellular GABA binding sites at the so-called β+/α− interfaces and a high affinity benzodiazepine binding site at the α+/γ− interface (see Figure 1a).3 A total of six α isoforms exist in mammalian species, along with three γ isoforms providing up to 18 high affinity sites.3 A similar modulatory site at the α+/β− interface (see Figure 1a) was described which differs from the benzodiazepine binding site only at the minus side that can be supplied by any of three β isoforms.4 As there are receptors that do not contain any γ− subunits,5 and thus also no benzodiazepine binding site, ligands of this site may target receptor populations partially distinct from those that bind benzodiazepines. Hence, such ligands are expected to display different and novel action profiles depending on the certain α and β subunits forming the α+/β− interfaces.6, 7 Moreover, some unwanted effects of ligands targeting the benzodiazepine binding site might be mediated by these sites. Together, an improved understanding of ligand recognition at the many related GABAA receptor extracellular interfaces is needed to improve rational drug development. While most benzodiazepines seem to be more selective for the αx+/γ2− (x = 1, 2, 3 and 5) interfaces compared to the α+/β− interfaces, flurazepam was shown to have a “null modulatory” low affinity site at α1+/β2−.8 Ligands of different chemotypes possessing higher liability to interact with both interfaces have been described. Specifically, CGS 20625 and CGS 9895, a pyrazolopyridinone and a pyrazoloquinolinone (PQ), can bind to both α+/γ− and α+/β− sites, but with different affinities.4,

9

By binding to the benzodiazepine site (α+/γ− interface) with high

affinity, they act as null modulators similar to flumazenil (Ro 15-1788), whereas binding to the α+/β− interface with low affinity results in positive allosteric modulation of GABAA receptors. Hence, the observation that ligands of different chemotypes can interact with the extracellular α+/γ− and α+/β− sites raises mechanistic questions. Do ligands that interact with both interfaces utilize identical interactions with the α+ pocket part, and thus use the same binding mode? Are the allosteric effects mediated by a conserved mechanism at these two interfaces? Clarification of these issues will lead to a better understanding of how allosteric ligands elicit modulatory effects. Rigid high affinity benzodiazepine site ligands such as pyrazoloquinolinones or betacarbolines were used traditionally as molecular rulers to derive ligand based pharmacophore 3 ACS Paragon Plus Environment


Page 4 of 27

models that provide quantitative spatial rules for the interaction of a ligand with its respective binding site.10 The alignments of more flexible ligands, such as 1,4-benzodiazepines, with the rigid ligands yielded pharmacophore models which identify putative interaction patterns with the protein for different chemotypes that share a binding site. While originally four possible alignments of benzodiazepines with a pharmacophore model based on high affinity ligands from several chemotypes were considered equivalent,10 only one was adapted into a popular unified pharmacophore model.11 Later, when the first protein structure data of homologues became available, computational docking of individual ligands started to augment the pharmacophore approach. In 2012, Richter et al. utilized computational docking to study the binding modes of benzodiazepine derivatives like diazepam and flumazenil (Ro 15-1788) at the α1+/γ2− interface.12 The assumption of a highly similar (i.e. common) mutual binding mode for ligands from closely related chemotypes was integrated in the workflow for the evaluation of binding modes10,

12

and yielded three

candidate “common” binding modes (CBM I, II and III) for benzodiazepine ligands. Of these, binding mode one (BM I) was found for many benzodiazepine ligands to be most consistent with a large body of experimental data. Virtual screening into BM I complexes subsequently led to the discovery of novel ligands.12 Next, binding mode two (BM II) was proposed to be the correct one for diazepam based on a combination of cysteine reactive diazepam derivatives with mutational analysis and a similar virtual screening approach as the one employed in the earlier work.13 Together these studies lead to controversial models for diazepam binding, and additionally open the question whether all benzodiazepines indeed utilize a common binding mode.11,

13

Here,

using a combined computational and experimental approach we challenge the notion of “common binding modes” for the group of ligands with a 1,4-benzodiazepine core. In the present study, we examined the related benzodiazepine site at the α1+/γ2− interface, as well as the low affinity α1+/β3− site and their interactions with 1,4-benzodiazepines and with ligands from the pyrazoloquinolinone and pyrazolopyridinone classes.

RESULTS AND DISCUSSION Choice and generation of the conversion mutants We reasoned that the incorporation of those amino acid residues of the γ2− side that promote high affinity binding of benzodiazepine site ligands into the β3− subunit (conversion mutants) ought to enable benzodiazepine binding and should increase the affinity of promiscuous ligands such as pyrazoloquinolinones that bind to the α1+/β3− site with lower affinity. Thus, the 4 ACS Paragon Plus Environment

Page 5 of 27 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


differences between the γ2− and the β3− sides were compared and the surfaces of the minus subunits that interact with the ligand were examined in our models14 based on the recently released β3 homopentamer− crystal structure15. Figure 1b shows the relevant conserved area of the crystal structure that is defined by a total of eight amino acid residues on the segments D, G and E. Only four of these residues are different between γ2− and β3− (compare Figure 1c and Figure 1d), and all four are known determinants for benzodiazepine site ligand recognition. Therefore, we consecutively introduced these four vicinal amino acids (two on segment D, one on segment G, and one on segment E) into the β3− side (see Figure 1d, blue oval) which resulted in a continuous patch of γ2-like pocket surface area with which the ligand can interact (see Figure 1c, d, purple oval).

Figure 1: a) Schematic view of the extracellular domain of a GABAA receptor. The GABA binding sites are labeled with “GABA”, the high affinity benzodiazepine/ CGS site with “ha”, and the low affinity CGS site with “la”. b) View of the binding site “through” the plus side (yellow thin tube) onto the minus side (blue). Binding site forming segments and loops are labeled by Arabic letters (D, E, F and G). Amino acids unique to the γ2 subunit are highlighted in blue whereas amino acids that are conserved between γ2 and β3 are highlighted in purple. c, d) Detailed view of the γ2 (c) and β3 (d) minus sides. Conserved amino 5 ACS Paragon Plus Environment


Page 6 of 27

acids are highlighted in purple whereas unique amino acids are highlighted in either blue (γ2− side) or red (β3− side). The blue oval is highlighting the four amino acids selected for mutational analysis (d) and the purple oval is indicating the entire “identical surface” that results from the four conversion mutants. Note that segment (loop) F of the γ2 subunit may differ considerably due to low conservation. The order in which we introduced the four point mutations was based in part on published findings, and in part on geometric considerations. Position γ2Ala79 (corresponding to β3Gln64, Figure 1c, d) of the loop D was proposed to be needed for high affinity binding of several imidazobenzodiazepines16 in a mutational study, where larger side chains in this position interfered with binding of these ligands.16 Thus, we reasoned that we first generate space that seems to be needed for ligand access or binding by replacing the bulkier β3Gln64 with the homologous alanine in the mutant β3Q64A. All other mutations were made on the background of this construct and chosen to systematically increase the γ2-like surface area by consecutive conversions. Thus, we next converted β3Tyr62 into the homologous phenylalanine, generating the construct β3Q64A;Y62F. Moving from segment D to segment G, we generated the triple mutant β3Q64A;Y62F;D43Y. We continued to increase the mutated surface further by introducing the G127T conversion17 resulting in β3Q64A;Y62F;D43Y;G127T (see Figure 2a, b and Supplementary Figures S1 and S2).9 Short names for the constructs are introduced in Table 1. Mutations on the so-called loop F would further increase the converted surface. We thus tried to introduce mutations in loop F, but we observed a reduction of functional expression with the increase of mutated position in the oocyte expression system as well as in HEK293 cells.18 Moreover, attempts to convert short F loop pieces with two different chimeras showed only low expression both in oocytes and in HEK 293 cells, and thus could not be characterized (see Supplementary Figure S3).18 Since loop F extensively participates in stabilizing minus side structure and conformation,15 conversion mutations on this segment may thus lead to problems with protein folding that would manifest in reduced functional protein. We thus discontinued further conversions and characterized the four mutations for which GABA currents were sufficient (see Supplementary Figure S2).


Page 7 of 27 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Table 1: Summary of the conversion mutants and the naming convention. Short name

Mutations

α1β3D1

α1β3Q64A

α1β3D2

α1β3Q64A;Y62F

α1β3D2G

α1β3Q64A;Y62F;D43Y

α1β3D2GE

α1β3Q64A;Y62F;D43Y;G127T



Page 8 of 27

The mutations increase the potency of CGS ligands We hypothesized that the stepwise conversion of β3 into γ2 should be paralleled by successive changes of the receptor’s response to compounds that interact with α1+/γ2− with high affinity, and with α1+/β3− with low affinity. We thus characterized modulation of GABA EC3 currents by the pyrazoloquinolinone CGS 9895 and the pyrazolopyridinone CGS 20625 in α1β3 WT and α1β3(mut) receptors for each construct.

Figure 2: Concentration-response curves of GABA EC3 modulation by CGS 9895 (a) and CGS 20625 (b) in wild type α1β3 (*) and α1β3γ2 () receptors and the mutated α1β3D1 (), α1β3D2 (), α1β3D2G () and α1β3D2GE () receptors with EC50 values: for CGS9895 23 μM (*), 38 μM (), 0.6 μM (), 1 μM (), 0.3 μM (), 0.03 μM () and for CGS 20625 20 μM (*), 35 μM (), 3 μM (), 10 μM (), 2 μM () and 0.3 μM (), respectively. The EC50 values are estimated by fitting the data to a hill slope of 1, since in all but α1β3D2GE saturation was not reached. Data represent mean ± SEM (n=3-11). CGS 9895 and CGS 20625 are low potency modulators at α1β3 and α1β3γ2 receptors, and the concentration-response curves are identical for these receptors, indicating that the effect of these ligands are exclusively mediated via the low affinity α1+/β3− site and not by the high affinity benzodiazepine site at which the compounds bind as high affinity silent or “null” modulators.4, 9 In the single mutated α1β3D1 receptor we observe a ~40-fold left shift of the EC50 for CGS 9895 and a ~7-fold left shift for CGS 20625 (see Figure 2a, b) compared to the wild type. The double mutated α1β3D2 receptors exhibit similar potency and at low concentrations also comparable efficacy, thus, the change from β3Tyr62 to phenylalanine makes almost no difference for the potency of these compounds, while efficacy increased. The third mutation leads to a further slight left shift of the dose response curve of CGS 20625 measured with the triple mutant α1β3D2G receptor (see Figure 2b), but not for CGS 9895 (see Figure 2a), while efficacy changes differentially for the two compounds. The fourth mutation chiefly causes a decrease of the maximum efficacy for both compounds (see Figure 2a, b), in line with the null 8 ACS Paragon Plus Environment

Page 9 of 27 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


modulatory effect observed at the benzodiazepine pocket. This seems to suggest that position Thr142 in the γ2 subunit and the homologous Gly127 in β3 drive the null (silent) modulatory effect at the benzodiazepine site and the positive modulation these compounds can exert at the α1+/β3− interfaces.4 For CGS 9895, a small positive modulation of the quadruple mutant α1β3D2GE is already seen at 100 nM compound concentration, indicating a further increase in potency (see Figure 2a). The mutations generally impact on efficacy, where both increases and decreases are observed. Similar effects have been observed previously; efficacy generally appears to be very easily affected by subtle changes in protein-ligand interactions.7, 17

Ro 15-1788 and Ro 15-8670 act as positive allosteric modulators in α1β3(mut) receptors α1β3 Having demonstrated that the mutated pockets show increased affinity for the CGS compounds, we moved on to test benzodiazepines. The classical benzodiazepine antagonist flumazenil (Ro 15-1788) is a high affinity null modulator at the benzodiazepine site in most subtypes,19 like CGS 9895. Indeed, we observed a positive modulation similar to that of the CGS compounds in the mutants (see Figure 3a). In contrast, the imidazobenzodiazepine Ro 15-8670 does not modulate the GABA-induced current in α1β3D1 and α1β3D2 receptors. However, in the α1β3D2G and α1β3D2GE receptors, both flumazenil and Ro 15-8670 act as positive modulators (see Figure 3b). The aromatic residue in loop G (γ2Y58) that we introduce in the third mutation thus seems to be necessary for activity of Ro 15-8670. As for the CGS compounds, the fourth mutation induces a reduction in the maximum efficacy that we observe, whereby the modulatory effect remains much stronger for the two tested imidazobenzodiazepines. Additional radioligand binding studies were attempted but proved challenging due to low expression of the quadruple mutant in HEK293 cells (see Supplementary Figures S3 and S4).



Page 10 of 27

Figure 3: Dose response curves of GABA EC3 modulation by Ro 15-1788 (a) and Ro 15-8670 (b) in wild type α1β3(*) and α1β3γ2 () receptors and the mutated α1β3D1 (), α1β3D2 (), α1β3D2G () and α1β3D2GE () receptors with EC50 values: for Ro 15-1788 17 μM (), 82 μM (), 6 μM (), 0.2 μM () and for Ro15-8670 0.2 μM (), 16 μM () and 1 μM (). The EC50 values are estimated by fitting the data to a hill slope of 1, since in many cases saturation was not reached. The respective EC50 values were low or no efficacy was observed could not be obtained, namely for Ro 15-1788 in wild type receptors (α1β3, α1β3γ2) and for Ro 15-8670 in the mutated receptors (α1β3D1 and α1β3D2). Data represent mean ± SEM (n=1-6). Effect of GABA EC3 modulation by 10 μM Diazepam (c) in wild type α1β3 and α1β3γ2 receptors and the mutated α1β3D1, α1β3D2, α1β3D2G and α1β3D2GE receptors. Data represent mean ± SEM (n=2-4). Sample traces of electrophysiological recordings can be found in Supplementary Figure S5.

The classical 1,4 benzodiazepine diazepam, interestingly, is inactive in all of our mutated receptors (see Figure 3c). This finding is rather surprising because diazepam displays excellent overlap with Ro 15-8670 in all common binding mode hypotheses12 as well as in all pharmacophore models of the high affinity benzodiazepine binding site.10 Moreover, diazepam is the smallest compound of the series that we tested here (total hydrophobic surface area: diazepam = 411.27, flumazenil = 518.90, Ro 15-8670 = 618.24), and thus would be predicted to 10 ACS Paragon Plus Environment

Page 11 of 27 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


readily fit into the binding site volume that is utilized by Ro 15-8670 - if these ligands were to use a common binding mode.

Computational docking suggests distinct binding modes for certain benzodiazepines We performed computational docking in search for a structural hypothesis for our experimental findings. In order to compare with our previous work, the same docking protocol as in Middendorp et al. 2014 was employed.13 Here, we docked into models based on the closer homologue, the β3 homopentameric structure 4COF.15 Overall, docking into models based on 4COF produced very similar pose spaces as those seen in the previous studies where AChBP structures were utilized as model templates.12, 13 In the new models the benzodiazepine core also can assume our previously observed binding modes BM I and BM II as reported by Richter et al. for the high affinity α1+/γ2− site, but not equally represented among the three different ligands. The largest ligand (Ro 15-8670) displayed among the top 20 ranked poses (ranked by ChemScore fitness function as implemented in GOLD20) only BM I orientations (7 times BM I) in the wild type α1+/γ2− site as well as in the α1+/β3D2GE− mutant (7 times BM I) and no BM II orientations (see Figure 4a, c). For flumazenil (Ro 15-1788) we observed among the top 20 ranked poses 4 times the BM I orientation in the wild type as well as in the mutant construct, suggesting a common binding mode with Ro 15-8670. Closer inspection of the BM I pose of Ro 15-8670 reveals that the pendant phenyl ring and the tyrosine we introduced in the third mutation make hydrophobic interactions and can also be stacked, which permits stabilizing pi-pi interactions (see Figure 4e). This yields a convincing structural hypothesis for the lack of activity in the single and double mutants. The charged β3D43 would make strong repulsive interactions with the aromatic ring of the ligand in this position. Thus, for flumazenil and Ro 15-8670 computational docking produces our previously published BM I as the most convincing candidate, in agreement with Richter et al. 2012.12 As in the previous study, a wide range of sparsely populated additional poses is observed as well.12 In contrast to the results with the imidazobenzodiazepines, docking of diazepam into the wild type receptor revealed poses in both orientations, BM I and BM II, but the poses of BM II were consistently ranked higher than the poses in BM I (among the top 10 ranked poses 4 times BM II and 2 times BM I. Thus, our computational findings in the wild type site, based on 4COF, are in agreement with the previous work of Middendorp et al. (based on an AChBP template) indicating



Page 12 of 27

BM II to be indeed a favorable binding mode for diazepam in the high affinity α1+/γ2− binding site.13 Interestingly, the pose space of diazepam in the α1+/β3D2GE− mutant did not provide any convincing binding pose and scores were generally low. Thus, the computational docking suggests that there is no favorable bound state complex in the quadruple mutant with diazepam, which is fully in agreement with the observed absence of effects of diazepam in the α1+/β3D2GE− mutant (see Figure 3c). A detailed analysis of our models revealed a smaller pocket volume in the α1+/β3−D2GE pocket due to a conformationally restrained large sidechain compared to the α1+/γ2− pocket (see Figure 4f and Supplementary Figure S6). If we enforce a BM II position of diazepam we thus observe severe clashes of the BM II bound diazepam in the pocket (see Figures 4d, f and Supplementary Figure S6) that would be incompatible with binding - consistent with the lack of activity in our experiments. On the contrary, BM I would sterically fit into the mutant pocket, but these poses feature no convincing interactions and thus low scores. BM I features, in the case of the imidazobenzodiazepines, strong interactions between the minus side and the five-membered ring (see Supplementary Figure S7). Lack of the five-membered ring thus seems to prevent diazepam from forming a stable BM I complex. It is important to note that the two binding modes BM I and BM II are pseudosymmetrical because the two lipophilic, aromatic rings position in the pocket and are related to each other nearly as mirror images (see Supplementary Figure S8). Shape complementarity with the pocket and thus binding will be strongly driven by the additional substituents on the ligand core. In summary, these results are in accordance with our biological data and strongly suggest that different benzodiazepine site ligands do not necessarily exhibit a common binding mode but rather can feature completely distinct binding modes. Thus, bound state structures should be considered for individual ligands in this case and not for the entire chemotype.


Page 13 of 27 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 4: General orientation of BM I and BM II of the benzodiazepine core scaffold and distinct binding modes of Ro 15-8670 and diazepam in the α1β3D2GE mutant. Color codes for the protein ribbon: α1 subunit: yellow; γ2 subunit: blue, β3 subunit: red, conversion mutation sites: blue. a) BM I of the benzodiazepine core scaffold at the α1+/γ2− interface. The fused phenyl ring (highlighted in orange) is directed towards the α1 subunit while the pendant phenyl ring (highlighted in purple) is oriented towards the γ2 subunit. Binding site forming segments (Strands and loops A-F) are labeled by Arabic letters; b) BM II of the benzodiazepine core scaffold at the α1+/γ2− interface. The fused phenyl ring (highlighted in orange) is directed towards the γ2 subunit while the pendant phenyl ring (highlighted in purple) is oriented towards the α1 subunit. Labels as in panel a; c) BM I of Ro 15-8670 (α1 subunit, yellow; β3D2GE mutant, red); d) BM II of diazepam (α1 subunit, yellow; β3D2GE mutant, red) is sterically hindered by a severe clash between the annealed ligand fragment and loop C tip, as well as β3D43Y. The steric clashes are schematically indicated by orange asterisks. e) BM I of Ro 15-8670 (α1 subunit, yellow; β3D2GE mutant, red) in the quadruple mutant (left half), where D43Y is in a position allowing hydrophobic and potentially pi-pi interactions with the ligand’s pendant phenyl ring and in comparison superimposed with a model with the native D43 (right half). The acidic side chain in close proximity to the phenyl ring would cause strong repulsive interactions, preventing binding.; f) Rotameric states of Y58 in the wild type pocket (left) and D43Y in the conformationally restrained β3D2GE mutant (right). 13 ACS Paragon Plus Environment


With a combined in silico and mutational approach we investigated the interactions of three benzodiazepine ligands with their high affinity α1+/γ2− binding site and with a series of engineered α1+/β3(mut) sites. To probe the molecular determinants required for high affinity binding at the α1+/γ2− interface, we sought to successively convert the binding pocket containing the low affinity β3 minus side into one with sufficient γ2-like properties to display high affinity benzodiazepine recognition. Ligands of the pyrazoloquinolinone type that interact with both wild type interfaces displayed changes in apparent affinity in the mutants consistent with a stepwise conversion from the β3-like into a γ2-like pattern. Subsequently, three representative benzodiazepines were tested with the same constructs. While flumazenil (Ro 15-1788) is a nonbinder at the α1+/β3− interface, the first mutation already enables flumazenil (Ro 15-1788) to interact with α1β3Q64A as low potency positive allosteric modulator. This mode of action is very similar to the interaction of pyrazoloquinolinones which show comparable positive allosteric modulation at the α1+/β3− interface that occurs with low potency.4, 9 Furthermore, this result is consistent with a previously proposed binding orientation of imidazobenzodiazepines, where space is required in position γ2Ala79.16 The triple and quadruple mutants induced further left shifts in the concentration-response curves of flumazenil (Ro 15-1788), suggesting a continuous increase in favourable interactions. Intriguingly, the phenyl substituted analogue Ro 15-8670 also modulates the α1β3(mut) receptors starting at conversion mutation 3, and also in mutation 4 (see Figure 3b). Here, the loop G position (that introduces γ2Tyr58 into the β3 subunit) enables interaction of this ligand with the mutated principal subunit. The structural hypothesis from the docking study suggests a pi-pi stacking of the ligand’s pendant phenyl ring with the tyrosine. Thus, the β3 subunit bearing the quadruple mutations provides an engineered flumazenil binding site with properties resembling the γ2 subunit. Paradoxically, diazepam remains completely inactive in the same engineered binding site. The computational docking study resolves the seemingly paradox experimental findings by indicating that ligands of the imidazobenzodiazepine-type (such as Ro 15-8670 or flumazenil) utilize a binding mode previously termed BM I. In this binding orientation, the fused core fragment of the ligand is sandwiched between the two subunits and the ester group on the imidazole ring points, in agreement with previous findings, towards γ2Ala79 on segment D.16 In contrast, diazepam seems to utilize BM II where the fused ring system is mainly in contact with the minus side forming subunit, as has been suggested previously.13 Residual differences between the γ2 minus side and our β3D2GE quadruple mutant are irrelevant for BM I, but incompatible with BM II where a larger and more rigid part of the ligand is in contact with the minus side. The lack of activity of diazepam in our mutated constructs, together with the docking 14 ACS Paragon Plus Environment

Page 14 of 27

Page 15 of 27 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


results that favour BM II poses for it, are not proving BM II usage and also do not directly reject a possible BM I usage. However, BM II usage is consistent with previous work13 as well as the most plausible hypothesis emerging from our data. The notion that benzodiazepines might utilize multiple binding modes has been discussed previously. The approach by Zhang et al 1995 was based on rigid and planar ligands utilized in pharmacophore model building.10 Four alignments of benzodiazepines with these “ruler ligands” (i.e. four benzodiazepine binding modes) have originally been proposed as equivalent solutions.10 Guided

by

our

data

we

now

propose

a

limited

common

binding

mode

for

12

imidazobenzodiazepine type ligands (BM I), while diazepam possesses a distinct one (BM II).13 Generalizations to other ligands should be carefully validated on a per ligand case. In

the

accompanying

manuscript21

the

binding

modes

of

chemically

related

imidazobenzodiazepine and benzodiazepine positive modulators with substituents that induce chirality are investigated in the wild type αx+/γ2− (x = 1, 2, 3 and 5) high affinity pockets. There we also present direct evidence for usage of BM I by a chiral imidazobenzodiazepine, and for BM II by chiral diazepam-analogues in the native pocket. These converging and independent lines of evidence in combination support the proposed distinct binding modes. Furthermore, our data suggest a highly conserved mechanism of both ligand recognition and allosteric modulation at the two investigated interfaces. First, introduction of successive amino acids that impart γ2-like character to the β3 subunit’s binding site introduced increasing apparent affinity of several chemically distinct ligands to the engineered binding site. At the same time, the same mutations, that all are localized directly in the binding site, enabled modulation of the α1β3(mut)

receptors

by

several

positive

modulators

of

imidazobenzodiazepine,

pyrazoloquinolinone and pyrazolopyridinone chemotypes. This implies that the entire mechanism that elicits positive modulation downstream of the ligand recognition site is equivalent in the β3 subunit compared to the γ2 subunit. Variations in ligand efficacy among the individual mutants are relatively small, and also suggest that the positive modulation elicited by interface binding sites is a conserved theme slightly varied by the exact shape of each interface.



Page 16 of 27

METHODS GABAA receptor subunits and mutated subunits cDNA´s of rat GABAA receptor subunits α1, β3 and γ2S were cloned as described.22 For the cloning of mutated β3 subunits, pCI expression vectors (Promega) were used. Four mutant constructs were designed by the method of “gene SOEing” PCR technique23 using a β3pCI vector as template, which had a the full length of a rat β3 subunit (1421bp) incorporated in a pCI vector

24

. The mutant products replaced parts of the wild type sequence on the vector. The

Escherichia Coli strain XL1-Blue (Stratagene) was used for plasmid amplification and isolation. Transformations were performed as published previously.25 The wild type β3pCI vector, as well as the resulting mutant products, were digested with the restriction enzymes PstI and XhoI (Promega) for subsequent ligation. Thus, a total of six different mutated β3pCI constructs were generated. The first construct β3D1pCI carries the mutation Y62F. Here, the amino acid residue Tyr62 (of the mature peptide) in the β3 subunit was changed to the homologous amino acid residue Phe of the γ2 subunit. Construct β3D2pCI carries the mutations Y62F and Q64A, β3DGpCI the mutations Y62F, Q64A and D43Y and β3DGEpCI the mutations Y62F, Q64A, D43Y and G127T (see Supplementary Figure S1). Construct β3DGE-F7pCI carries the mutations Y62F, Q64A, D43Y, G127T and a chimeric sequence containing the seven amino acids γ2(186-192), and β3DGE-F12pCI the mutations Y62F, Q64A, D43Y, G127T and a chimeric sequence containing the twelve amino acids γ2(186-197) (see Supplementary Figure S1). All constructs were verified by sequencing. Wild-type or mutated β3 subunits were then co-expressed together with a wild type α1 subunit, to form a functional pentameric α1β3 GABAA receptor.26 RNA Preparation In vitro transcription of mRNA was based on the cDNA expression vectors encoding for rat GABAA receptor subunits α1, β3, γ2 and the four β mutants (β3D1, β3D1, β3DG and β3DGE).19 After linearizing the cDNA vectors with appropriate restriction endonucleases, the cDNA was purified and concentrated with the DNA Clean and ContentratorTM Kit (Zymoresearch, Catalog No. D4005). Capped transcripts of the purified cDNA were produced using the mMESSAGE mMACHINE® T7 transcription kit (Ambion) and polyadenylated using the Ambion PolyA tailing kit (Ambion). After transcription and polyadenylation the RNA was purified with the MEGAclearTM Kit (Ambion, Catalog No. AM1908). The final RNA concentration was measured on NanoDrop® 16 ACS Paragon Plus Environment

Page 17 of 27 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


ND-1000 and finally diluted and stored in diethylpyrocarbonate-treated water at -80°C. For the microinjection, the RNA of αβ receptor combinations was mixed at 1:1 ratio (which leads to αβ receptors that consist of predominantly 3 beta and 2 alpha subunits).27 All receptor combinations had a final concentration of 56 ng µl-1.

Two electrode voltage clamp (TEV) in Xenopus laevis oocytes Mature female Xenopus laevis (Nasco) were anesthetized and oocytes were isolated and prepared as described in Simeone et al. 2017.7 Healthy defolliculated oocytes were injected with an aqueous solution of mRNA. A total of 4.5 ng of mRNA per oocyte was injected with a Nanoject II (Drummond). After injection of mRNA, oocytes were incubated at 18 °C (ND96 + antibiotic) for 2-3 days for αβ receptors and for 3-4 days for αβγ receptors before recording. When cells were measured at later time points, oocytes were stored at +4 °C instead of 18 °C. Electrophysiological recordings were performed as specified in Simeone et al. 2017.7 Data were analysed using GraphPad Prism v.6. and plotted as concentration-response curves, as defined in Simeone et al. 2017.7 Homology modelling and molecular docking

Homology models of the extracellular α1+/γ2− interface (benzodiazepine site) and the α1+/β3− interface (CGS binding site) were made based on 4COF15 using Modeller28 as described previously.14 The quadruple mutant was created by introduction of single point mutations (β3Q64A; Y62F; D43Y; G127T) using MOE. Molecular Docking was performed using GOLD.20 The putative binding sites were defined by a cut-off distance of 11.5 Å around the residue α1Ser204 of the C-loop of the α1 subunit. Further, we selected ten amino acids with flexible side chains (for α1+/γ2−: γ2Tyr58, γ2Phe77, γ2Thr142, α1His101, α1Tyr159, α1Val202, α1Ser204, α1Ser205, α1Thr206 and α1Tyr209; for α1+/β3−DGE: β3D43Y, β3Y62F, β3G127T, α1His101, α1Tyr159, α1Val202, α1Ser204, α1Ser205, α1Thr206 and α1Tyr209) and set a soft potential to increase to backbone flexibility of the C-loop (α1Ser204, α1Ser205, α1Thr206 and α1Gly207) (see Supplementary Figure S9). To ensure convergence of the sampling, 100 genetic algorithm runs were performed using diazepam, flumazenil (Ro 15-1788) and Ro 15-8670. The ligands were built in MOE using the M conformation of the seven-membered ring that is supported by experimental studies as described previously.12 Before the docking run the ligands 17 ACS Paragon Plus Environment


Page 18 of 27

were energetically minimized using the MMFF94 force field. The 100 poses per ligand were rescored with the ChemScore fitness function implemented in GOLD.


Page 19 of 27 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Investigated compounds Compounds were purchased from the following sources: [N-Methyl-3H]Ro 15-1788 (87.0 Ci/mmol) (Perkin-Elmer Life Sciences), diazepam (7-Chloro-1,3-dihydro-1-methyl-5-phenyl-2H1,4-benzodiazepin-2-one), flumazenil (Ro 15-1788) (Ethyl-8-fluoro-5,6-dihydro-5-methyl-6-oxo4H-imidazo[1,5-a][1,4]benzodiazepine-3-carboxylate), Ro 15-8670 (Ethyl 8-chloro-6-phenyl-4Hbenzo[f]imidazo[1,5-a][1,4]diazepine-3-carboxylate) and CGS 20625 (2-(4-Methoxyphenyl)5,6,7,8,9,10-hexahydrocyclohepta[b]pyrazolo[3,4-d]pyridin-3(2H)-one) (Tocris Bioscience). ASSOCIATED CONTENT Supporting Information: Supporting Figure S1, alignment of mutated β3pCI constructs in „loops“ G, D, E and F; Supporting Figure S2, GABA concentration-response curves; Supporting Figure S3, cell surface expression of GABAA receptors containing wild-type α1 and β3 or mutated β3 subunits; Supporting Figure S4, saturation assay of [3H]Ro 15-1788 binding in HEK cell membranes of wild type α1β3γ2 and recombinant α1β3DGE receptors and potency of CGS 20625 for inhibition of [3H]Ro15-1788 binding in HEK cell membranes of recombinant α1β3DGE receptors; Supporting Figure S5, sample traces of electrophysiological recordings; Supporting Figure S6, the α1β3D2GE pocket is sterically more crowded than the α1γ2 pocket; Supporting Figure S7, represententative BM I poses of Ro 15-8670 and diazepam in comparison; Supporting Figure S8, pseudosymmetry of benzodiazepine core; Supporting Figure S9, representative docking results of Figure 4; Supplementary Methods. The Supporting information is available free of charge via the Internet at http://pubs.acs.org. AUTHOR INFORMATION Corresponding Author *E-mail: [email protected] Author Contributions §

These authors contributed equally to this work.

NOTES The authors declare no competing financial interest. 19 ACS Paragon Plus Environment


Page 20 of 27

ACKNOWLEDGMENTS We gratefully thank A. Dereky for some measurements. This project is funded by FWF grants W1232 MolTag and P27746.


Page 21 of 27 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


REFERENCES

1. 2.

3. 4.

5. 6.

7.

8.

9.

10.

11.

12.

13.

14.

Chebib, M., and Johnston, G. A. (2000) GABA-Activated ligand gated ion channels: medicinal chemistry and molecular biology, J. Med. Chem. 43, 1427-1447. Olsen, R. W., and Sieghart, W. (2008) International Union of Pharmacology. LXX. Subtypes of gamma-aminobutyric acid(A) receptors: classification on the basis of subunit composition, pharmacology, and function. Update, Pharmacol. Rev. 60, 243-260. Sieghart, W. (2015) Allosteric modulation of GABAA receptors via multiple drugbinding sites, Adv. Pharmacol. (San Diego, CA, U.S.) 72, 53-96. Ramerstorfer, J., Furtmüller, R., Sarto-Jackson, I., Varagic, Z., Sieghart, W., and Ernst, M. (2011) The GABA(A) Receptor α+β− Interface: A Novel Target for Subtype Selective Drugs, J. Neurosci. 31, 870-877. Mortensen, M., and Smart, T. G. (2006) Extrasynaptic alphabeta subunit GABAA receptors on rat hippocampal pyramidal neurons, J. Physiol. 577, 841-856. Sieghart, W., Ramerstorfer, J., Sarto-Jackson, I., Varagic, Z., and Ernst, M. (2012) A novel GABA(A) receptor pharmacology: drugs interacting with the α(+)β(−) interface, Br. J. Pharmacol. 166, 476-485. Simeone, X., Siebert, D. C. B., Bampali, K., Varagic, Z., Treven, M., Rehman, S., Pyszkowski, J., Holzinger, R., Steudle, F., Scholze, P., Mihovilovic, M. D., Schnürch, M., and Ernst, M. (2017) Molecular tools for GABAA receptors: High affinity ligands for β1containing subtypes, Sci. Rep. 7, 5674. Baur, R., Tan, K. R., Luscher, B. P., Gonthier, A., Goeldner, M., and Sigel, E. (2008) Covalent modification of GABAA receptor isoforms by a diazepam analogue provides evidence for a novel benzodiazepine binding site that prevents modulation by these drugs, J. Neurochem. 106, 2353-2363. Varagic, Z., Wimmer, L., Schnurch, M., Mihovilovic, M. D., Huang, S., Rallapalli, S., Cook, J. M., Mirheydari, P., Ecker, G. F., Sieghart, W., and Ernst, M. (2013) Identification of novel positive allosteric modulators and null modulators at the GABAA receptor alpha+beta- interface, Br. J. Pharmacol. 169, 371-383. Zhang, P., Zhang, W., Liu, R., Harris, B., Skolnick, P., and Cook, J. M. (1995) Synthesis of novel imidazobenzodiazepines as probes of the pharmacophore for "diazepaminsensitive" GABAA receptors, J. Med. Chem. 38, 1679-1688. Clayton, T., Chen, J. L., Ernst, M., Richter, L., Cromer, B. A., Morton, C. J., Ng, H., Kaczorowski, C. C., Helmstetter, F. J., Furtmuller, R., Ecker, G., Parker, M. W., Sieghart, W., and Cook, J. M. (2007) An updated unified pharmacophore model of the benzodiazepine binding site on gamma-aminobutyric acid(a) receptors: correlation with comparative models, Curr. Med. Chem. 14, 2755-2775. Richter, L., de Graaf, C., Sieghart, W., Varagic, Z., Mörzinger, M., de Esch, I. J. P., Ecker, G. F., and Ernst, M. (2012) Diazepam-bound GABA(A) receptor models identify new benzodiazepine binding-site ligands, Nat. Chem. Biol. 8, 455-464. Middendorp, S. J., Puthenkalam, R., Baur, R., Ernst, M., and Sigel, E. (2014) Accelerated discovery of novel benzodiazepine ligands by experiment-guided virtual screening, ACS Chem. Biol. 9, 1854-1859. Puthenkalam, R., Hieckel, M., Simeone, X., Suwattanasophon, C., Feldbauer, R. V., Ecker, G. F., and Ernst, M. (2016) Structural Studies of GABA-A receptor binding sites: Which experimental structure tells us what?, Front. Mol. Neurosci. 9. 21 ACS Paragon Plus Environment


15. 16.

17.

18.

19.

20. 21.

22.

23.

24.

25. 26. 27.

28.

Page 22 of 27

Miller, P. S., and Aricescu, A. R. (2014) Crystal structure of a human GABAA receptor, Nature 512, 270-275. Kucken, A. M., Teissere, J. A., Seffinga-Clark, J., Wagner, D. A., and Czajkowski, C. (2003) Structural requirements for imidazobenzodiazepine binding to GABA(A) receptors, Mol. Pharmacol. 63, 289-296. Mihic, S. J., Whiting, P. J., Klein, R. L., Wafford, K. A., and Harris, R. A. (1994) A single amino acid of the human gamma-aminobutyric acid type A receptor gamma 2 subunit determines benzodiazepine efficacy, J. Biol. Chem. 269, 32768-32773. Dereky, A. L. (2010) Engineering of a Benzodiazepine binding site into GABAA receptors composed of alpha1 and beta3 subunits, In Master Thesis, University of Vienna, Vienna. Ramerstorfer, J., Furtmuller, R., Vogel, E., Huck, S., and Sieghart, W. (2010) The point mutation gamma 2F77I changes the potency and efficacy of benzodiazepine site ligands in different GABAA receptor subtypes, Eur. J. Pharmacol. 636, 18-27. Jones, G., Willett, P., Glen, R. C., Leach, A. R., and Taylor, R. (1997) Development and validation of a genetic algorithm for flexible docking, J. Mol. Biol. 267, 727-748. Elgarf, A. A., Siebert, D. C. B., Steudle, F., Draxler, A., Li, G., Huang, S., Cook, J. M., Ernst, M., and Scholze, P. (2018) Different Benzodiazepines Bind with Distinct Binding Modes to GABAA Receptors, ACS Chem. Biol. Ebert, V., Scholze, P., and Sieghart, W. (1996) Extensive heterogeneity of recombinant gamma-aminobutyric acidA receptors expressed in alpha 4 beta 3 gamma 2-transfected human embryonic kidney 293 cells, Neuropharmacology 35, 1323-1330. Horton, R. M. (1993) In Vitro Recombination and Mutagenesis of DNA, In PCR Protocols: Current Methods and Applications (White, B. A., Ed.), pp 251-261, Humana Press, Totowa, New Jersey. Ehya, N., Sarto, I., Wabnegger, L., and Sieghart, W. (2003) Identification of an amino acid sequence within GABA(A) receptor beta3 subunits that is important for receptor assembly, J. Neurochem. 84, 127-135. Fuchs, K., Zezula, J., Slany, A., and Sieghart, W. (1995) Endogenous [3H]flunitrazepam binding in human embryonic kidney cell line 293, Eur. J. Pharmacol. 289, 87-95. Baumann, S. W., Baur, R., and Sigel, E. (2001) Subunit arrangement of gammaaminobutyric acid type A receptors, J. Biol. Chem. 276, 36275-36280. Che Has, A. T., Absalom, N., van Nieuwenhuijzen, P. S., Clarkson, A. N., Ahring, P. K., and Chebib, M. (2016) Zolpidem is a potent stoichiometry-selective modulator of α1β3 GABAA receptors: evidence of a novel benzodiazepine site in the α1-α1 interface, Sci. Rep. 6, 28674. Sali, A., and Blundell, T. L. (1993) Comparative protein modelling by satisfaction of spatial restraints, J. Mol. Biol. 234, 779-815.


Page 23 of 27 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Content Figure 166x70mm (150 x 150 DPI)

ACS Paragon Plus Environment


Figure 1 139x123mm (150 x 150 DPI)


Page 24 of 27

Page 25 of 27 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 2 177x56mm (300 x 300 DPI)



Figure 3 167x112mm (300 x 300 DPI)


Page 26 of 27

Page 27 of 27 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 4 165x219mm (300 x 300 DPI)


Engineered Flumazenil Recognition Site Provides ... - ACS Publications

Recommend Documents