International Journal of Molecular Sciences

Article GABAA Receptor Ligands Often Interact with Binding Sites in the Transmembrane Domain and in the Extracellular Domain—Can the Promiscuity Code Be Cracked?

Maria Teresa Iorio 1, Florian Daniel Vogel 2, Filip Koniuszewski 2, Petra Scholze 3 , Sabah Rehman 2, Xenia Simeone 2, Michael Schnürch 1, Marko D. Mihovilovic 1 and Margot Ernst 2,* 1 Institute of Applied Synthetic Chemistry, TU Wien, Getreidemarkt 9/163-OC, 1060 Vienna, Austria; [email protected] (M.T.I.); [email protected] (M.S.); [email protected] (M.D.M.) 2 Department of Molecular Neurosciences, Center for Brain Research, Medical University of Vienna, Spitalgasse 4, 1090 Vienna, Austria; fl[email protected] (F.D.V.); fi[email protected] (F.K.); [email protected] (S.R.); [email protected] (X.S.) 3 Department of Pathobiology of the Nervous System, Center for Brain Research, Medical University of Vienna, Spitalgasse 4, 1090 Vienna, Austria; [email protected] * Correspondence: [email protected]; Tel.: +43-1-40160-34065



Received: 29 November 2019; Accepted: 26 December 2019; Published: 3 January 2020


Abstract: Many allosteric binding sites that modulate gamma aminobutyric acid (GABA) effects have been described in heteropentameric GABA type A (GABAA) receptors, among them sites for benzodiazepines, pyrazoloquinolinones and etomidate. Diazepam not only binds at the high affinity extracellular “canonical” site, but also at sites in the transmembrane domain. Many ligands of the benzodiazepine binding site interact also with homologous sites in the extracellular domain, among them the pyrazoloquinolinones that exert modulation at extracellular α+/β sites. Additional − interaction of this chemotype with the sites for etomidate has also been described. We have recently described a new indole-based scaffold with pharmacophore features highly similar to pyrazoloquinolinones as a novel class of GABAA receptor modulators. Contrary to what the pharmacophore overlap suggests, the ligand presented here behaves very differently from the identically substituted pyrazoloquinolinone. Structural evidence demonstrates that small changes in pharmacophore features can induce radical changes in ligand binding properties. Analysis of published data reveals that many chemotypes display a strong tendency to interact promiscuously with binding sites in the transmembrane domain and others in the extracellular domain of the same receptor. Further structural investigations of this phenomenon should enable a more targeted path to less promiscuous ligands, potentially reducing side effect liabilities.

Keywords: receptors; GABAA; benzodiazepines (BZ); binding sites; etomidate

1. Introduction

GABAA receptors are pentameric anion channels heavily expressed in the mammalian central nervous system. They are formed of five homologous subunits, encoded by 19 different genes. Despite the significant subtype heterogeneity, it is commonly accepted that the majority of receptors is formed of two α and β subunits and a single γ subunit. The subunits form a pentameric GABA gated ion pore, with subunit interfaces formed by a principal (or +) and a complementary (or ) face of each subunit as −

Int. J. Mol. Sci. 2020, 21, 334; doi:10.3390/ijms21010334 www.mdpi.com/journal/ijms Int.Int. J. J. Mol. Mol. Sci. Sci. 20202020, ,2121, ,334 334 2 2of of 20 20 of each subunit as depicted in Figure 1. Treatments of epilepsy, anxiety states and sleep disorders, as welldepicted as anesthetics, in Figure1 .target Treatments GABA ofA receptors epilepsy, anxiety[1,2]. Among states andthese sleep drugs, disorders, benzodiazepines as well as anesthetics,are widely prescribedtarget GABA andA receptorsare known [1 to,2 ].exert Among their these effects drugs, via a benzodiazepinesbinding site on the are extracellular widely prescribed domain and of the are receptor.known to Specifically, exert their e fftheects binding via a binding site of site benzodiazepines on the extracellular is situated domain at of the the extracellular receptor. Specifically, domain the binding site of benzodiazepines is situated at the extracellular domain (ECD)- interface α+/γ (ECD)- interface α+/γ− (site 1, Figure 1) and homologous with the GABA binding site at the ECD-− (site 1, Figure1) and homologous with the GABA binding site at the ECD- β+/α interface [3,4]. β+/α− interface [3,4]. − BenzodiazepinesBenzodiazepines interact interact additionally additionally with with another another low low affinity affinity binding binding site site within within the the transmembranetransmembrane domain domain (TMD), (TMD), at at the the TMD-interface β+/+/α−α (site(site 3) 3) [5]. [5]. This This binding binding site site is is chiefly chiefly − knownknown as as the the site site of of action action for for the the clinically clinically used used drug drug etomidate, etomidate, an andd also also one one of of the the interaction interaction sites sites forfor propofol propofol [6,7]. [6,7]. BesidesBesides the the clinically clinically relevant relevant drugs, drugs, a a wide wide range range of of compounds compounds have have been been identified identified as as modulatorsmodulators of of the the GABA GABAAA receptorsreceptors [1,8–10]. [1,8–10].

FigureFigure 1. (a(a) )Schematic Schematic rendering rendering of of a a GABAA receptor.receptor. The The extracellular extracellular domain domain (ECD) (ECD) and and transmembranetransmembrane domain domain (TMD) (TMD) contai containn various various binding binding sites sites [11]. [11]. The The binding binding sites sites studied studied in in this this workwork are atat interfacesinterfaces formed formed by by a a principal principal (+ (+)) subunit subunit face face and and a complementary a complementary ( ) subunit(−) subunit face. face. The − Theshape shape resembles resembles vaguely vaguely a space a fillingspace renderingfilling rendering of the protein’s of the ECDprotein’s and TMD,ECD whileand TMD, intracellular while intracellulardomain (ICD) domain shape is(ICD) a purely shape schematic is a purely rendering schematic based rendering on more based remote on homologues.more remote homologues. (b) Schematic (planesb) Schematic through planes the ECDthrough (top) the and ECD TMD (top) (bottom) and TMD of (bottom) a canonical of a canonicalαβγ receptor, αβγ receptor, the GABA the sites GABA are indicated as dark grey ellipses. Site 1 (blue), the high affinity benzodiazepine site, is at the ECD– α+/γ sites are indicated as dark grey ellipses. Site 1 (blue), the high affinity benzodiazepine site, is at the− ECD–interface; α+/γ− site interface; 2 (blue), site which 2 (blue), confers which modulatory confers emoduffectslatory of pyrazoloquinolinone effects of pyrazoloquinolinone ligands is at the ligands ECD– α+/β interface, and site 3 (orange), the etomidate site [7] occurs twice and is located at the TMD– is at the− ECD– α+/β− interface, and site 3 (orange), the etomidate site [7] occurs twice and is located at β+/α interfaces below the GABA binding sites. the TMD–− β+/α− interfaces below the GABA binding sites.

InIn a a previous previous work, work, we we identified identified a a series series of of indole indole derivatives, derivatives, sharing sharing pharmacophore pharmacophore features features with pyrazoloquinolinones (PQs), as GABA modulators (Figure2a) [ 12]. PQs are known to bind with with pyrazoloquinolinones (PQs), as GABAA A modulators (Figure 2a) [12]. PQs are known to bind withhigh high affinity affinity at the at BZthe siteBZ site (site (site 1) and1) and to to modulate modulate the the receptors receptors via via the the interactioninteraction withwith a second second binding site of the ECD, at the interface between α+ and β (site 2) [13]. An interaction of PQs with binding site of the ECD, at the interface between α+ and β−− (site 2) [13]. An interaction of PQs with sitesite 3 3 has has also also been been reported reported [14] [14] For For this this study, study, we selected the most efficacious efficacious compound of of the the previouslypreviously published library, MTI163 (Figure2 2).).

Int. J. Mol. Sci. 2020, 21, 334 3 of 20 Int. J. Mol. Sci. 2020, 21, 334 3 of 20

FigureFigure 2. (a 2.) Structure(a) Structure of MTI163; of MTI163; (b) structure(b) structure of DCBS192; of DCBS192; (c) pharmacophores (c) pharmacophores of indole of indole derivatives derivatives (left)(left) and and of PQs of PQs (right); (right); (d) superposition(d) superposition of pharmacophores of pharmacophores of PQ of andPQ and indole indole derivatives. derivatives. Panels Panels c c andand d depict d depict the pharmacophoresthe pharmacophores already already discussed discussed in [12 in]. [12]. The high pharmacophoric overlap of the indole derivatives in comparison with the PQs (Figure2c,d) The high pharmacophoric overlap of the indole derivatives in comparison with the PQs (Figure suggests similar binding behavior. We therefore considered the three possible binding sites as object 2c,d) suggests similar binding behavior. We therefore considered the three possible binding sites as of our studies: site 1 (ECD- α+/γ ) site 2 (ECD- α+/β ), and site 3 (TMD- β+/α ). To investigate object of our studies: site 1 (ECD-− α+/γ−) site 2 (ECD-− α+/β−), and site 3 (TMD- −β+/α−). To investigate the putative binding and mechanism of action of the new scaffold at these three sites, we performed the putative binding and mechanism of action of the new scaffold at these three sites, we performed functional and mutational studies, as well as radioligand displacement assays. Computational docking functional and mutational studies, as well as radioligand displacement assays. Computational studies complement the experimental findings in the context of the recently published cryo EM docking studies complement the experimental findings in the context of the recently published cryo structure with the PDB identifier 6HUP [4], which features diazepam in binding sites 1 and 3. EM structure with the PDB identifier 6HUP [4], which features diazepam in binding sites 1 and 3. 2. Results 2. Results 2.1. Chemistry 2.1. Chemistry The indole derivative MTI163 was synthesized as previously described [12]. DCBS192 was synthesizedThe indole according derivative to MTI163 a previously was synthesiz reporteded route as previously [13,15]. For described details of[12]. the DCBS192 synthesis, was seesynthesized AppendixA. according to a previously reported route [13,15]. For details of the synthesis, see Appendix A. 2.2. Pharmacological Assessments 2.2. Pharmacological Assessments 2.2.1. Radioligand Displacement Assays 2.2.1.To investigate Radioligand a possible Displacement binding Assays at site 1, where PQs are binding with high affinity, we performed displacementTo investigate assays with a possible [3H]-flunitrazepam binding at (Figuresite 1, 3wherea,b). In PQs contrast are binding to the pyrazoloquinolinone with high affinity, we DCBS192,performed which displacement displaced the radioligandassays with already [3H]-flunitrazepam at picomolar concentrations (Figure 3a,b). (Ki =In0.33 contrast0.07 nMto ),the ± MTI163pyrazoloquinolinone did not displace DCBS192, [3H]-flunitrazepam, which displaced even the ra atdioligand high µM already concentrations at picomolar (leading concentrations to a Ki of(Ki> 30= 0.33µM). ± 0.07 nM), MTI163 did not displace [3H]-flunitrazepam, even at high µM concentrations (leadingDespite to the a Ki similarity of >30µM). of the pharmacophoric features, these results exclude a binding of the indole derivativeDespite the MTI163 similarity at site of 1.the It pharmacophoric seems surprising fe atatures, first sightthese thatresults the exclude high pharmacophore a binding of the similarityindole isderivative not reflected MTI163 in similar at site binding 1. It seems properties. surprising However, at first we sight have that previously the high observed pharmacophore that interactionssimilarity with is not site reflected 1 do not in follow similar pharmacophore binding properties. features However, [16], and we recently have previously published observed structures that confirminteractions that even with ligands site 1 do of not the follow benzodiazepine pharmacophore chemotype features do [16], not and display recently binding published modes structures with confirm that even ligands of the benzodiazepine chemotype do not display binding modes with

Int. J. Mol. Sci. 2020, 21, 334 4 of 20 Int. J. Mol. Sci. 2020, 21, 334 4 of 20 alignedaligned pharmacophore pharmacophore features. features. Specifically, Specifically, flumazenil flumazenil (PDB (PDB structure structure 6D6T 6D6T [17]) [17]) and and alprazolam alprazolam (PDB(PDB structure structure 6HUO 6HUO [4]) [4]) bind bind at at the the high high affinity affinity site site with with completely completely di differentfferent binding binding modes modes and and nono overlap overlap of of pharmacophore pharmacophore features features (Figure 33c,d).c,d).

Figure 3. Radioligand displacement assay of [3H]-Flunitrazepam by DCBS192 (a) and MTI163 (b). 3 FigureMembranes 3. Radioligand from rat cerebellumdisplacement were assay incubated of [ H]-Flunitrazepam with 2 nM [3H]-flunitrazepam by DCBS192 (a) in and the MTI163 presence (b of). 3 Membranesvarious concentrations from rat cerebellum of the two were ligands. incubated Data shownwith 2 arenM the[ H]-flunitrazepam mean SEM of in three the independentpresence of ± variousexperiments concentrations performed of in the duplicates two ligands. each. (Datac,d) Superpositionshown are the of mean the flumazenil ± SEM of boundthree independent 6D6T (green) experimentsand the alprazolam performed bound in 6HUOduplicates (yellow) each. in (c two,d) diSuperpositionfferent perspectives of the flumazenil demonstrate bound that the 6D6T structurally (green) andrelated the moleculesalprazolam do bound not bind 6HUO with aligned(yellow) pharmacophore in two different features perspectives at all. Thedemonstrate heteroatoms that of the the structurallymolecules are related depicted molecules in different do colors:not bind blue-nitrogen, with aligned red-oxygen, pharmacophore green-chloro. features (e) Structuresat all. The of heteroatomsalprazolam andof the flumazenil. molecules are depicted in different colors: blue-nitrogen, red-oxygen, green- chloro. (e) Structures of alprazolam and flumazenil. 2.2.2. Functional and Mutational Studies, Beta Isoform Profile 2.2.2. Functional and Mutational Studies, Beta Isoform Profile In the next step, we aimed to compare the modulation of MTI163 in wild type receptors α1β1, α1β2,Inα the1β 3next and step, a concatenated we aimed versionto compare of α1 theβ3γ mo2, termeddulationα1 ofβ3 MTI163γ2cct (see in methods)wild type GABA receptorsA receptors α1β1, α(Figure1β2, α41a).β3 and MTI163 a concatenated modulates GABA-elicitedversion of α1β3 currentsγ2, termed in α1β32γ and2cct α(see1β3 methods) similarly. GABA In contrast,A receptorsα1β1 (Figurereceptors 4a). show MTI163 a much modulates lower e ffiGABA-elicitedcacy of MTI163 currents modulation in α1β (Figure2 and 4αa).1β3 In similarly. contrast toIn MTI163,contrast, which α1β1 receptorsmodulates showα1β a3 much with lower veryhigh efficacy effi cacy,of MTI163 DCBS192 modulation hardly (Figure modulates 4a). In at contrast all (130 to MTI163,14% at 30 whichµM). ± modulatesThe‘very low α1β effi3 withcacy ofvery this high substance efficacy, precluded DCBS192 a morehardly detailed modulates functional at all (130 analysis ± 14% and at also30 µM). indicates The verythat thelow high efficacy similarity of this ofsubstance pharmacophore precluded features a more does detailed not translate functional into analysis similar and pharmacology also indicates for thatthese the two high compounds. similarity of pharmacophore features does not translate into similar pharmacology for these two compounds.

Int. J. Mol. Sci. 2020, 21, 334 5 of 20 Int. J. Mol. Sci. 2020, 21, 334 5 of 20

Figure 4. (a) Concentration-dependentConcentration-dependent MT MTI163I163 modulation of GABA EC3-5 elicitedelicited current current in in α1β1,1, α1β2,2, αα11ββ33 and and αα11ββ33γγ2cct.Response2cct. Response at ateach each point point of me of measurementasurement is represented is represented as mean± as mean SEMSEM (n = ± 3–4).(n = 3–4 Response). Response at 10µM at 10 isµ highlightedM is highlighted in red incolor. red color.The y-axis The yis-axis broken is broken at indicated at indicated values valuesfor better for visualizationbetter visualization of all curves of all curves in the inpanel. the panel.For better For bettervisualization, visualization, a full ascale full version scale version of the of α1 theβ1 αdose1β1

responsedose response curve curveis in the is inAppendix the Appendix A (FigureA (Figure A5). (b A5) Comparison). ( b) Comparison of MTI163 of modulation MTI163 modulation of GABA of3-5 elicitedGABA3-5 currentelicited in current α1β1, α in1βα2,1β α1,1βα2N265S,1β2, α1β α2N265S,1β3, andα α1β1β3,3N41R and α1 receptorsβ3N41R receptorsat 10µM. atBars 10 µareM. given Bars are as given as mean SEM (n = 4–8). Each individual data point is displayed by a dot. Statistically significant mean±SEM (n± = 4–8). Each individual data point is displayed by a dot. Statistically significant differencesdifferences were determined by one-way ANOVA wi withth Tukey’s multiple comparisons post-hoc test, test, *** corresponds toto p < 0.0001. α β α β While modulation in the binary subunit combinations α1β2 and α1β33 does not reach saturation µ α β γ α β up to 3030 µM,M, α11β33γ2cct2cct andand α11β11 receptors receptors show show a typicala typical sigmoidal sigmoidal dose dose response response curve curve (Figure (Figure4a). β 4a).The observedThe observedisoform β isoform profile profile could reflectcould reflect interactions interactions with either with site either 2 at thesite extracellular2 at the extracellular domain’s α+/β interface or site 3 at the transmembrane domain’s β+/α interface. We thus utilized previously domain’s− α+/β− interface or site 3 at the transmembrane domain’s− β+/α− interface. We thus utilized β previouslypublished mutations published [ 18mutations,19] to investigate [18,19] to theirinvestigate impact their on modulation. impact on modulation.3N41R is aβ partial3N41R conversionis a partial β β β β β conversionof the ECD of site the of ECD3 into site the of β31 into [18], the while β1 [18],2N265S while is β a2N265S conversion is a conversion at site 3 of theat site2 3 into of the1, β which2 into βis1, known which tois reverseknown theto reverse effects the of etomidate effects of etomidate and loreclezole and loreclezole [20,21]. [20,21]. While the mutation β3N41R did not affect the modulation of MTI163 (α1β3N41R: 1032 189% vs. While the mutation β3N41R did not affect the modulation of MTI163 (α1β3N41R: 1032± ± 189% 1110 109% in α1β3, Figure4b), the mutation β2N265S resulted in a drop in modulation comparable vs. 1110± ± 109% in α1β3, Figure 4b), the mutation β2N265S resulted in a drop in modulation to the one obtained in α1β1 (α1β2: 1108 68% vs. α1β2N265S: 198 26% vs. α1β1 158 12%, comparable to the one obtained in α1β1 (α1±β2: 1108 ± 68% vs. α1β2N265S:± 198 ± 26% vs. α1β1± 158 ± Figure4b).These results suggest that the extracellular α+/β site (site 2) doesn’t contribute to the 12%, Figure 4b).These results suggest that the extracellular α+/−β− site (site 2) doesn’t contribute to the β modulatory eeffectffect of MTI163, whereas thethe interactioninteraction withwith N265N265 locatedlocated atat thethe β++ side of site 3 in the TMD seems to be crucial for the modulation. 2.3. Structural Hypothesis/Computational Docking 2.3. Structural Hypothesis/Computational Docking Since all the experimental findings strongly suggest that MTI163 exerts the modulatory effect Since all the experimental findings strongly suggest that MTI163 exerts the modulatory effect via the “etomidate site” at the TMD β+/α site involving β2N265S, we performed computational via the “etomidate site” at the TMD β+/α−− site involving β2N265S, we performed computational docking to investigate the possible binding of MTI163 to the equivalent site of 6HUP which harbors docking to investigate the possible binding of MTI163 to the equivalent site of 6HUP which harbors diazepam. A redocking of diazepam was performed first to test whether the protocol we use recovers diazepam. A redocking of diazepam was performed first to test whether the protocol we use recovers the experimental structure [16]. Poses with very high overlap to the one observed in the cryo-EM the experimental structure [16]. Poses with very high overlap to the one observed in the cryo-EM structure (6HUP [4]) were indeed found within the top 10 ranked docking results, evaluated with two structure (6HUP [4]) were indeed found within the top 10 ranked docking results, evaluated with scoring functions as well in the absence or presence of flexible sidechains (see Methods). We thus two scoring functions as well in the absence or presence of flexible sidechains (see Methods). We thus proceeded to investigate MTI163 and etomidate. proceeded to investigate MTI163 and etomidate. MTI163 was used for the computational investigation in its nonionized form since it was found MTI163 was used for the computational investigation in its nonionized form since it was found experimentally that this acid cannot be deprotonated even under strongly basic conditions (2N NaOH). experimentally that this acid cannot be deprotonated even under strongly basic conditions (2N Hence, deprotonation under physiological conditions is highly unlikely. This is also supported by NaOH). Hence, deprotonation under physiological conditions is highly unlikely. This is also density functional theory (DFT) calculations, which show that an intramolecular hydrogen bond supported by density functional theory (DFT) calculations, which show that an intramolecular stabilizes the molecule by 11.2 kcal/mol (see AppendixA, Figure A1), providing an explanation for hydrogen bond stabilizes the molecule by 11.2 kcal/mol (see Appendix A, Figure A1), providing an explanation for the low acidity observed. Additionally, our previously reported crystal structure [12]

Int. J. Mol. Sci. 2020, 21, 334 6 of 20

Int. J. Mol. Sci. 2020, 21, 334 6 of 20 the low acidity observed. Additionally, our previously reported crystal structure [12] displays the moleculedisplays inthe the molecule same conformation in the same asconformation calculated as as the calculated most stable as the one most and hencestable thisone conformer and hence was this usedconformer for docking. was used For etomidate,for docking. we haveFor etomidate, no indication we tohave select no a particularindication conformer,to select a it particular was thus dockedconformer, as a fullyit was flexible thus docked ligand. as a fully flexible ligand. DockingDocking ofof bothboth ligandsligands resultsresults in in a a broad broad diversity diversity of of highly highly scored scored poses. poses. Etomidate Etomidate docking docking posesposes display display high high overlap overlap with with thethe diazepamdiazepam bound bound statestateand and manymany posesposes featurefeatureinteractions interactions withwith aminoamino acids acids known known to to impact impact on theon the ligand’s ligand’s potency potency and eandfficacy efficacy (Appendix (AppendixA, Figure A, A2Figure). To furtherA2). To disentanglefurther disentangle etomidate’s etomidate’s very diverse very posing diverse space posi wouldng space require would going require to considerably going to higher considerably level of theory,higher orlevel to haveof theory, some or experimental to have some data experimental on its active data conformation, on its active whichconformation, is out of which scope is for out this of study.scope Thefor this posing study. space The accessible posing tospace MTI163 accessible was more to MTI163 limited, was but stillmore featured limited, several but still alternative featured solutionsseveral alternative among the solutions top 10 posesamong in the the top two 10 scoringposes in functions the two scoring that were functions used herethat were (Appendix used hereA, Figure(Appendix A3). InA, termsFigure of A3). a consensus In terms ofof thea consensus top 10 poses, of the one top featured 10 poses, a prominent one featured interaction a prominent with N265,interaction and thus with it N265, represents and thus a highly it represents plausible a highly candidate plausible for the candidate MTI163 boundfor the MTI163 state in thisbound site, state as shownin this insite, Figure as shown5. Docking in Figure into 5. the Do correspondingcking into the sitecorresponding at the β1+ interfacesite at the revealed β1+ interface a very revealed similarly a diversevery similarly posing spacediverse (Appendix posing spAace, Figure (Appendix A4). A, Figure A4).

FigureFigure 5. 5.Representative Representative dockingdocking posepose ofof MTI163MTI163 inin thethe diazepamdiazepam pocketpocket ofof 6HUP.6HUP. (a(a)) The The MTI163 MTI163 pose,pose, which which is is fully fully consistent consistent with with all all experimental experimental evidence, evidence, is is displayed displayed with with the the pocket-forming pocket-forming aminoamino acids acids that that were were set set flexible flexible during during docking. docking. (b(b)) Overlay Overlay of of diazepam diazepam (yellow) (yellow) and and the the MTI163 MTI163 (red)(red) pose pose in in 6HUP 6HUP ligands ligands structures structures (top), (top), ligands ligands surfaces surfaces (bottom). (bottom). TheThe leftleft subunitsubunit (pink)(pink) isisβ β3,3, thethe right right one one (blue) (blue) is isα α1.1.

WeWe proceededproceeded toto examineexamine highlyhighly rankedranked posesposes afterafter energyenergy minimizationminimization toto obtainobtain furtherfurther insight.insight. AsAs expected,expected, energyenergy minimizationminimization improvedimproved thethe ligand–proteinligand–protein interactions.interactions. However,However, thethe minimizedminimized poses poses derived derived from from consensus consensus top 10top poses 10 po stillses could still notcould be robustlynot be robustly ranked with ranked standard with post-dockingstandard post-docking methods at methods a modest at level a modest of theory. level Thus, of theory. the overall Thus, conclusion the overall from conclusion computational from docking is that the diazepam site at the TMD β3+/α1 site from 6HUP can readily accommodate computational docking is that the diazepam site at the− TMD β3+/α1− site from 6HUP can readily etomidateaccommodate and MTI163,etomidate and and all MTI163, three ligands and all interact three ligand with thes interact same set with of corethe same residues set of that core have residues been shownthat have to bebeen relevant shown for to thebe relevant site’s modulatory for the site’s eff ectsmodulatory (Figure5 effectsb). The (Figure orientation 5b). The in theorientation pocket is in the pocket is very different for diazepam and MTI163, where MTI163 poses are only in a confined space parallel to the helices that form the pocket, whereas diazepam (experimental structure) and etomidate (docking results) occupy a niche that is perpendicular to the interface. While the interacting

Int. J. Mol. Sci. 2020, 21, 334 7 of 20 very different for diazepam and MTI163, where MTI163 poses are only in a confined space parallel to the helices that form the pocket, whereas diazepam (experimental structure) and etomidate (docking Int. J. Mol. Sci. 2020, 21, 334 7 of 20 results) occupy a niche that is perpendicular to the interface. While the interacting sidechains are largelysidechains the sameare largely for all the three same ligands, for all thethree diff ligands,erent spatial the different orientation spatial of MTI163 orientation is not of suggestiveMTI163 is not for anysuggestive common for binding any common motif thatbinding generally motif drivesthat generally ligand recognition drives ligand in thisrecognition pocket. in this pocket. 2.4. Ligand Analysis 2.4. Ligand Analysis In a further step, we sought to obtain an overview concerning the frequency with which ligands In a further step, we sought to obtain an overview concerning the frequency with which ligands have been reported to interact with multiple distinct sites on a single pentameric receptor complex, have been reported to interact with multiple distinct sites on a single pentameric receptor complex, here limited to sites 1, 2 and 3, in αβγ subtypes. An exhaustive search of the literature resulted in here limited to sites 1, 2 and 3, in αβγ subtypes. An exhaustive search of the literature resulted in a a pool of ligands that are known for interactions with either only one, or any two, or all three sites pool of ligands that are known for interactions with either only one, or any two, or all three sites studied here as depicted in Figure6. studied here as depicted in Figure 6. For two of these binding sites, presumably selective compounds can be found, e.g., zolpidem For two of these binding sites, presumably selective compounds can be found, e.g., zolpidem for for the ECD site 1 and loreclezole [20,22] and tracazolate [1,23] for the TMD site 3. In contrast, for the ECD site 1 and loreclezole [20,22] and tracazolate [1,23] for the TMD site 3. In contrast, for site 2 site 2 in the ECD, no exclusive ligand has been identified so far. It also needs to be kept in mind that in the ECD, no exclusive ligand has been identified so far. It also needs to be kept in mind that interactions with site 2 may have gone unnoticed in the past because this site has been described in interactions with site 2 may have gone unnoticed in the past because this site has been described in 2011 and to this date, no radioligand has been published for it. 2011 and to this date, no radioligand has been published for it. Well-characterized showcases for promiscuous ligands are e.g., diazepam [5] and the β-Carboline Well-characterized showcases for promiscuous ligands are e.g., diazepam [5] and the β- DMCM [19,24], which have been described to bind at sites 1 and 3, while flurazepam binds at sites 1 Carboline DMCM [19,24], which have been described to bind at sites 1 and 3, while flurazepam binds and 2 (for the case of α1+β2 ). The pyrazoloquinolinones CGS 9895 and LAU 177, [14] as well as the at sites 1 and 2 (for the case− of α1+β2−). The pyrazoloquinolinones CGS 9895 and LAU 177, [14] as β-Carboline ZK91085 [19] were found to interact with all of the considered sites. well as the β-Carboline ZK91085 [19] were found to interact with all of the considered sites.

Figure 6. Graphical overview of chemical scascaffoldsffolds whichwhich bindbind toto distinctivedistinctiveinterfaces interfaces at atαβγ αβγ GABAGABAAA receptors.receptors. The The three three binding binding sites sites are are depicted depicted as colored as colored ellipses: ellipses: site 1, site ECD- 1, ECD-α+/γ−:α blue;+/γ site: blue; 2, ECD- site − 2,α+/ ECD-β−: orange;α+/β site: orange; 3, TMD- site β+/ 3,α− TMD- red. βSubstances+/α red. that Substances bind at more that bind than atone more site thanare placed one site at arethe − − placedintersection at the of intersection the different of thebindin differentg sites. binding All placements sites. All placementsrest on current rest onevidence current from evidence structural from structuralstudies [4,17] studies and [ 4decades,17] and decadesof mutational of mutational studies studies[19–21,23–26]. [19–21, 23Some–26]. of Some the ofplacements the placements are thus are thustentative. tentative. Of note, Of note,for both for boththe benzodiazepine the benzodiazepine site and site the and “etomidate the “etomidate site”, site”,many many more moreligands ligands have havebeen beendescribed, described, partly partly reviewed reviewed by Sieghart by Sieghart (2015) (2015) and Sieghart and Sieghart and Savic and Savic (2018) (2018) [1,10]; [1 ,however,10]; however, low lowaffinity affi nityinteractions interactions with with other other sites sites have have not notbeen been regularly regularly investigated. investigated.

In addition to the cases depicted in Figure 6, several other cases of ligands using at least two distinctive sites have been reported, but the assignment of the sites is not always completely clear. Flavonoids represent a big chemical group of natural compounds and their derivatives. Many of these have been shown to be high affinity ligands of site 1 [27]; however, modulatory effects of individual compounds have been shown to occur at different sites. A prominent example is the potent 6-methoxyflavone hispidulin [28,29]. Flavanol-3-esters have been found to mimic the binding

Int. J. Mol. Sci. 2020, 21, 334 8 of 20

In addition to the cases depicted in Figure6, several other cases of ligands using at least two distinctive sites have been reported, but the assignment of the sites is not always completely clear. Flavonoids represent a big chemical group of natural compounds and their derivatives. Many of these have been shown to be high affinity ligands of site 1 [27]; however, modulatory effects of individual compounds have been shown to occur at different sites. A prominent example is the potent 6-methoxyflavone hispidulin [28,29]. Flavanol-3-esters have been found to mimic the binding properties of loreclezole, since the positive modulatory effect of Fa131 on GABA elicited currents and was diminished by the mutations β2N265S and M236W, similarly to loreclezole/etomidate [25]. Another flavan-3-ol, Fa173, has been presented as specific antagonist of Fa131, loreclezole and etomidate. Low affinity potentiation of diazepam could also be blocked by Fa173, whereas high affinity diazepam effects, as well as modulation by neurosteroids or barbiturates were not affected—confirming the use of a common site by Fa173 and the selectively antagonized compounds [25]. Another example of a scaffold with individual ligands showing different degrees of promiscuity between the high affinity benzodiazepine site and other, unknown sites have been termed ROD compounds [30]. ROD compounds were derived from bicuculline, a GABA site antagonist, but turned out to be allosteric modulators and not bind at the GABA sites. In an effort to understand their mode of action, they were classified according to their pharmacological profile as R1 and R2 types. R1 substances bind to the benzodiazepine binding site (and completely inhibit binding of benzodiazepines) as well as another yet unknown site, named R1 site. In contrast, R2 compounds do not bind to the α+/γ − interface, nor do they inhibit benzodiazepine binding at all. The benzodiazepine effect is not additive for R1 compounds but is for R2s. Both classes were able to modulate αβ containing GABAA receptors, suggesting a distinct modulatory site for both classes that differs from the benzodiazepine binding site. Among compounds with a similar subtype profile as etomidate and loreclezole are several furanones (gamma-butyrolactones) [31]. For these, the lack of benzodiazepine site binding and sensitivity of modulatory efficacy to the β2N265S mutation were demonstrated, thus suggesting that they interact exclusively with site 3, similar to MTI163 [31]. An interesting question is whether common features determining either specificity for a particular site, or driving promiscuity, could be deduced and yield predictive structure activity rules. Given that diazepam and midazolam both interact with sites 1 and 3, while flurazepam does not interact with site 3, but interacts with sites 1 and 2, this task seems to be very complex and likely requires extensive study of focused libraries.

3. Discussion In our previous work [12], we published a library of novel modulators derived from the modification of the scaffold of pyrazoloquinolinones (Figure2). In this follow-up study, we aimed to investigate the binding sites of these novel structures. Due to the similarity of their pharmacophores, we hypothesized a binding behavior similar to the one of their parent compounds. PQs are known to bind with high affinity at the benzodiazepine binding site (site 1). More recently it was proven that they exert their modulation via a secondary binding site (site 2) [13], located between the α and β subunits in the ECD, where they bind at higher concentration. Additionally, binding at a TMD binding site (site 3) was suggested by mutational studies [14]. Here we investigated which of these binding sites are also used by MTI163 (Figure2) and aimed to compare it directly with the identically substituted pyrazoloquinolinone DCBS192. Site 1 has the advantage to allow a direct investigation of a possible binding, employing radioligands such as tritiated flunitrazepam. Since no displacement was observed for MTI163 (Figure3a), in contrast with the full displacement performed by DCBS192 (Figure3b), we excluded a possible binding at site 1. We then expanded our investigation of the pharmacological profile of MTI163, and investigated GABAA receptors containing different β subunits. A marked difference in efficacy exerted by MTI163 between β2/3 and β1 containing assemblies reminiscent of etomidate’s β2/3 profile prompted − − mutational analysis. Conversion mutants were employed, and strongly suggest that all of the Int. J. Mol. Sci. 2020, 21, 334 9 of 20 modulatory effect exerted by MTI163 originates from interactions with the binding site used by etomidate [20,26]. Taken together, all these experimental findings suggest a binding of MTI163 at site 3 only. The identically substituted pyrazoloquinolinone is a very weak modulator, and thus not accessible to functional/mutational analysis. For DCBS192 it remains thus unclear which sites it uses in addition to the high affinity binding at the benzodiazepine binding site. Thus, our experimental results indicated that the similarity in pharmacophore features between these two ligands did not at all translate into similar pharmacology. Our efforts to generate a rational hypothesis for this finding prompted a detailed analysis of available structural and pharmacological data. Given the observation that not only pyrazoloquinolinones display activity at multiple binding sites in GABAA receptors, we approached the phenomenon from the viewpoint of putative structure-activity rules (SAR) that may help to better predict for ligands their preferences for specific sites. We thus sorted known ligands into “bins” that reflect specific or promiscuous usage of the three binding sites under investigation as depicted in Figure6. Some of the assignments of ligands to the individual bins reflect the current status of the literature, but may have to be revised in future studies, as there still is a paucity of data on the ECD α+/β– site due to lacking radioligands or antagonists. This effort revealed some interesting observations: All chemotypes that have been described so far to interact with all three sites are relatively flat, rigid and aromatic structures—the β-Carbolines and the pyrazoloquinolinones. Of note, such scaffolds also tend to display extremely high affinity for site 1, as is the case for DCBS192 with its Ki of 0.33nM. This could be the case because the benzodiazepine binding site contains a number of aromatic sidechains that facilitate pi–pi stacking of aromatic rings. A particularly impressive case of extensive pi–pi stacking at a homologous site has been reported for acetylcholine binding protein, where a sandwich of three aromatic ligands occupies the binding site (PDB structure 4BFQ, [32]). However, the reverse conclusion that planar ligands are typical for site 1 is clearly not valid, and a very wide range of scaffolds has been described for this site [1,33]. Among the best known site 1 ligands are the benzodiazepines, for which many pharmacophore models have been developed in the past—all of which rested on the assumption that most or all benzodiazepines share a common binding mode in which common pharmacophore features overlap. Based on a mutational analysis combined with computational docking, we have recently challenged this notion [16], and more recent structures in fact confirmed that benzodiazepines do not generally share a common binding mode [4,17]. Thus, there is no single consensus pharmacophore applicable to the design of site 1 ligands. Benzodiazepines also have been demonstrated to interact with sites 2 and 3, where so far only isolated studies for small numbers of ligands have been performed and for most benzodiazepines it is unknown whether they are site 1 specific or not. Similarly, we asked whether it is possible to identify possible features that drive the interaction towards site 3. For this site, the diversity of scaffolds that was reported to elicit (sometimes β2/3 − preferring) modulatory effects is also vast, Figure6 provides several well-known examples, others are reviewed in Sieghart and Savic 2018 [1]. The only common set of features we identified was a combination of a hydrophobic feature and a hydrogen bond acceptor with a distance of from 5 to 6.5 A between these two descriptors, which is not enough to define any “core pharmacophore” for this site. Some first insights emerged from the diazepam bound 6HUP structure: In line with the large difference in affinity for the two sites (1 and 3), diazepam displays more and stronger interactions in site 1. In site 3, there are many branched hydrophobic amino acid sidechains that provide hydrophobic interactions, and can contribute with a multiplicity of rotameric states to induced fit phenomena. Moreover, this region of the receptor has been demonstrated to be highly flexible—thus, site 3 can be seen as a hydrophobic room that changes volume and shape both in response to conformational movements and in response to ligand binding events. It could be seen as a consequence that a wide variety of ligands, ranging from small molecules such as ethanol to rather large ones such as avermectin can bind with low to moderate affinity, and elicit strong modulatory effects due to the pocket position right at the hinge that couples the ECD with the TMD. Int. J. Mol. Sci. 2020, 21, 334 10 of 20

The 6HUP structure thus provides first structural insight into a phenomenon, which poses a serious challenge for rational drug development targeting GABAA receptors. While the extracellular interfaces display high subtype specificity, mostly owed to the large variable segment (Loop) C and other variable parts of the ECD, the interfaces at the upper TMD where etomidate is known to bind are highly conserved—featuring only a single variable amino acid that confers β isoform selectivity. Thus, compounds that should be subtype selective (site 1 ligands) show a high propensity for unwanted low affinity effects at a more conserved site, and thus limit the subtype selectivity. In spite of the considerable size of the problem, systematic studies are lacking, and thus large scale SAR investigations cannot be performed. Due to its highly flexible nature [11], the etomidate pocket can accommodate a surprising diversity of molecules, and thus, unwanted activity there needs to be monitored carefully at early stages in ligand development until more structural data accumulates that may allow in silico screening tools into appropriate models. Of final note, this phenomenon is not limited to the three sites discussed in the present work. Another intriguing example are bicuculline derivatives that originally were targeted at the GABA sites, many of which turned out to be modulatory ligands of the benzodiazepine site and additional unknown sites [30,34].

4. Materials and Methods

4.1. Compound Synthesis Organic solvents were purified when necessary by standard methods or purchased from commercial suppliers [35]. Chemicals were purchased from commercial suppliers and used without further purification. Thin-layer chromatography (TLC) was performed using silica gel and 60 aluminum plates containing fluorescent indicator from Merck, and detected either with UV light at 254 nm or by charring potassium permanganate (1 g KMnO4, 6.6 g K2CO3, 100 mg NaOH, 100 mL H2O in 1M NaOH, pH = 14) with heating. Flash column chromatography (FC) was carried out on a Büchi SepacoreTM MPLC system using silica gel 60 M (particle size 40–63 µm, 230–400 mesh ASTM, Macherey Nagel, Düren). Unless otherwise noted all compounds were purified with a ratio of 1/100 (weight (compound)/weight (silica)). Nuclear magnetic resonance (NMR) spectra were recorded on a Bruker Avance Ultrashield 400 (1H: 400 MHz, 13C: 101 MHz) and Bruker Avance IIIHD 600 spectrometer equipped with a Prodigy BBO cryo probe (1H: 600 MHz, 13C: 151MHz). The spectra are found in the supporting information (Supplementary Material), where chemical shifts are given in parts per million (ppm) and were calibrated with internal standards of deuterium labelled solvents CHCl3-d 1 13 1 13 ( H 7.26 ppm, C 77.16 ppm) and deuterated dimethyl sulfoxide DMSO-d6 ( H 2.50 ppm, C 39.52 ppm). Multiplicities are denoted by s (singlet), br s (broad singlet), d (doublet), dd (doublet of doublet), t (triplet), q (quartet), and m (multiplet). Melting points were determined with a Büchi Melting Point B-545 apparatus. High resolution mass spectrometry (HR-MS) was measured on an Agilent 6230 Liquid chromatography time-of-flight mass spectrometry (LC TOFMS) mass spectrometer equipped with an Agilent Dual AJS ESI-Source. (E)-3-((4-bromophenyl)diazenyl)-5-chloro-1H-indole-2carboxylic acid (MTI163). Compound MTI163 was synthesized according to a previously reported synthesis [12] in 60% 1 yield (red solid, 30 mg). M. p > 300 ◦C; H NMR (600 MHz, DMSO-d6) d 7.42 (dd, J = 8.7, 2.2 Hz, 1H), 7.55 (d, J = 8.7 Hz, 1H), 7.78 (d, J = 8.7 Hz, 2H), 7.86 (d, J = 8.6 Hz, 2H), 8.43 (d, J = 2.1 Hz, 1H), 12.86 13 (s, 1H); C NMR (151 MHz, DMSO-d6) d 114.8 (d), 118.2 (s), 122.5 (d), 123.4 (s), 123.9 (d, 2C0), 126.1 (d),128.2 (s),132.4 (d, 2C),133.4 (s),133.9 (s, 2C), 152.0 (s), 161.6 (s); HR-MS (ESI): 377.9636 [M + H]+; (calcd. 377.969). 2-(4-bromophenyl)-8-chloro-2,5-dihydro-3H-pyrazolo[4,3-c]quinolin-3-one (DCBS192). Compound DCBS192was synthesized according to reported routes in 90% yield (yellow solid, 1 1g). [36] M. p. decomposition > 300 ◦C; H NMR (400 MHz, DMSO-d6) d 7.60–7.66 (m, 2H), 7.71–7.76 (m, 2H), 8.16–8.18 (m, 1H), 8.18–8.24 (m, 2H), 8.8 (d, J = 5.5 Hz, 1H), 12.99 (s, 1H).); 13C NMR (151 MHz, Int. J. Mol. Sci. 2020, 21, 334 11 of 20

DMSO-d6) d 106.1 (s), 116.1 (s), 119.9 (s), 120.3 (d), 121.2 (d), 121.8 (d), 130.4 (d), 130.8 (s), 131.6 (d), 134.3 (s), 139.2 (s), 140.0 (d), 142.4 (s), 161.6 (s); HR-MS (ESI): 373.9694 [M + H]+; (calcd. 373.9690). For synthesis of precursors see AppendixA1.

4.2. Electrophysiology in Xenopus Laevis Oocytes Preparation of mRNA for rat α1, β1, β2, β2N265S, β3, β3N41R, and γ2 subunits as well as α1β3γ2 concatenated constructs (α1β3γ2cct) and electrophysiological experiments with Xenopus laevis oocytes were performed as described previously [12,18,19]. α1β3γ2cct was prepared as described previously [37–39] and included subunits in the following order: (α1-β3-α1; γ2-β3). Oocytes were purchased from EcoCyte Bioscience (Dortmund, Germany). Cold oocytes were washed in Ca2+-free ND96 medium (96 mM NaCl, 2 mM KCl,1 mM MgCl2,5mM MHEPES; pH 7.5). Healthy, defolliculated cells were injected with an aqueous solution of mRNA (70 ng/µL). A total of 2.5 ng of mRNA per oocyte was injected. Subunit ratio was 1:1 for α1βx (x = 1, 2, 3, 3N41R) and 1:1 (α1-β3-α1; γ2-β3) for concatenated α1β3γ2 receptors. Injected oocytes were incubated for at least 36 h before electrophysiological recordings. Oocytes were placed on a nylon grid in a bath of NDE medium. For current measurements, oocytes were impaled with two microelectrodes, which were filled with 2 M KCl and had a resistance of 1–3 MΩ. The oocytes were constantly washed by a flow of 6 mL/min NDE that could be switched to NDE containing GABA and/or drugs. Drugs were diluted into NDE from DMSO solutions resulting in a final concentration of 0.1% DMSO. To test for modulation of GABA currents, a concentration of GABA, which elicited 3–5% in receptors of the respective maximum GABA-elicited current, was applied to the cell with increasing concentrations of compounds. All recordings were performed at room temperature at a holding potential of 60 mV using a TURBO TEC-03X npi two − electrode clamp (npi electronic GmbH, Tamm, Germany). Data were digitized, recorded and measured using an Axon Digidata1550 low-noise data acquisition system (Axon Instruments, UnionCity, CA, USA). Data acquisition was done using pCLAMP v.10.5 (Molecular Devices™, Sunnyvale, CA, USA) at a 500-Hz sampling rate. Enhancement of the chloride current was defined as (IGABA+Comp/IGABA) 100, where I is the current response in the presence of a given compound and GABA, × GABA+Comp and IGABA is the control GABA current. The normalized current enhancement data was then used to nH construct dose response curves with the formula Y = Maximum/(1 + (EC50/x) ), whereas EC50 is the concentration of the compound that increases the amplitude of the GABA-evoked current by 50%, and nH is the Hill coefficient. Data were analysed using GraphPadPrism v.8. for Windows (GraphPad Software, La Jolla, CA, USA, www.graphpad.com) and plotted as bar graphs or dose response curves. Data are given as mean SEM from at least three oocytes of two and more oocyte batches. ± 4.3. Displacement Assays Rat cerebellar membranes were prepared and radioligand binding assays were performed as described previously [18,40]. In brief, membrane pellets were incubated for 90 min at 4 ◦C in a total of 500 µL of a solution containing 50 mM Tris/citrate buffer, pH = 7.1, 150 mM NaCl and 2 nM [3H]-flunitrazepam in the absence or presence of either 5 µM diazepam (to determine unspecific binding) or various concentrations of DCBS192 or MTI163 (dissolved in DMSO, final DMSO-concentration 0.5%). Membranes were filtered through Whatman GF/B filters and washed twice with 4 mL of ice-cold 50 mM Tris/citrate buffer. Filters were transferred to scintillation vials and subjected to scintillation counting after the addition of 3 mL Rotiszint Eco plus liquid scintillation cocktail. Nonlinear regression analysis of the displacement curves used the equation: log(inhibitor) vs. response—variable slope ((logIC50-x) Hillslope) with Top = 100% and Bottom = 0% Y = 100/(1 + 10 × ). IC 50 values were converted to Ki values using the Cheng–Prusoff relationship [41] Ki = IC50/(1 + (S/KD) with S being the concentration of the radioligand (2 nM) and the KD values (4.8 nM) as described in Simeone et al., 2017 [18]. All analysis were performed using GraphPad Prism version 8.3.0 for Mac OS X (GraphPad Software, La Jolla, CA, USA, www.graphpad.com). Int. J. Mol. Sci. 2020, 21, 334 12 of 20

4.4. Computational Docking Molecular docking experiments were performed with “GOLD” [42]. The recently published cryo-EM structure 6HUP was used to take advantage of the ligand bound conformation in site 3. Standard protein preparation was used, where all hydrogen atoms were added to the protein [16]. The binding site within the respective pocket was defined according to the coordinates of the diazepam - N1 of the amide group with a sphere size of 12A. Ten pocket-forming amino acids were set flexible, specifically β3/1 Thr262, Asn/Ser265 and Arg269 of the TM2 and Met286 and Phe289 of the TM3, and for α1 Gln229, Leu232 and Met236 for the TM1 and Thr265 and Leu269 for the TM2. Docking of MTI163 and etomidate was performed with all ten amino acid sidechains set flexible. Etomidate was treated as fully flexible ligand, while MTI163 was restrained to the conformer, which was observed experimentally (AppendixA, Figure A1). For the redocking of diazepam, sidechains were also either flexible or stayed fixed as they are in 6HUP, and the ligand remained fixed. Scoring poses were primarily ranked with GoldScore and rescored with ChemScore. One-hundred GoldScore top-ranked poses were analyzed from each run. Poses that were among top 20 of the consensus pool were analyzed further. Energy minimization was performed using the MOE function “energy minimize”, selecting the Amber10:EHT force field, together with fixed hydrogens and charges, all other parameters were left at default values. Ligand interaction plots were generated, displaying the interaction of a respective docking pose of a ligand with the receptor.

5. Conclusions The switch from a Pyrazoloquinoline (PQ) scaffold towards an indole based body induces striking difference in the binding properties of our substances. PQs were shown to bind at three different binding sites: (1) ECD- α+/γ , (2) ECD- α+/β , and (3) TMD- β+/α , possibly also at other TMD − − − interfaces. Surprisingly, MTI163 does not bind to binding site 1, is unlikely to exert modulation via binding site 2, but modulates GABAA receptors chiefly or even exclusively through the transmembrane “etomidate” site 3. It is not obvious which properties lead to the high promiscuity of PQs or to the site 3 specificity of MTI163. Similar discrepancies in ligand-based structure activity relationships of other chemotypes have been observed previously, but have not yet been systematically studied. Several benzodiazepines also interact with site 3, flurazepam interacts with site 2, while alprazolam and flumazenil seem to be more specific to site 1 only. This particular type of ligand promiscuity is very likely to be missed in standard assays aimed at refining interactions with a single site. With this study, we hope to raise awareness and to prompt further studies dedicated to this theme.

Supplementary Materials: Supplementary materials can be found at http://www.mdpi.com/1422-0067/21/1/334/s1. Author Contributions: Conceptualization, M.D.M. and M.E.; methodology, M.D.M., M.S., M.T.I. and M.E.; software, F.K.; validation and formal analysis, M.D.M., M.S., M.T.I. and M.E.; investigation, M.T.I., F.D.V., S.R., P.S., X.S., F.K.; resources, M.E. and M.D.M.; data curation, M.E.; writing—original draft preparation, M.T.I.; writing—review and editing, M.T.I., F.D.V.,M.S., P.S., M.E.; visualization, M.T.I., F.K., X.S., F.D.V., P.S., M.S.; supervision, M.E., M.S., M.D.M.; project administration, M.E. and M.D.M.; funding acquisition, M.E., M.D.M. All authors have read and agree to the published version of the manuscript. Funding: This work was financially supported by Austrian Science Fund FWF grant W1232 (graduate school program MolTag). Acknowledgments: The constructs β2N265S and α1β3γ2_cct (concatemers) are a kind gift from Erwin Sigel. Open Access Funding by the Austrian Science Fund (FWF). Conflicts of Interest: The authors declare no conflicts of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results. Int. J. Mol. Sci. 2020, 21, 334 13 of 20

Abbreviations

BZ Benzodiazepines GABA Gamma-aminobutyric acid EC Effective Concentration ECD Extracellular domain ICD Intracellular domain TMD Transmembrane domain PQs Pyrazoloquinolinone TLC Thin-layer Chromatography Int. J. Mol. Sci. 2020, 21, 334 13 of 20 Appendix A Appendix A Appendix A.1. Compound Synthesis Appendix A.1. Compound Synthesis

Scheme A1. Reagents and conditions: ( (aa)) ethylpyruvate, ethylpyruvate, AcOH, AcOH, EtOH, EtOH, rf, rf, 2 2 h, h, 95%; 95%; ( (bb)) PPA, PPA, 110 °C,◦C, 5 h, 60%; (c) DMF, KK22CO3,, 0 0 °C,◦C, 30 30 min, min, 70%; 70%; ( (dd)) NaOH NaOH (2N), (2N), rf, rf, 3 3 h, h, 60%. 60%.

Appendix A.1.1. Synthesis of ethyl-(Z)- ethyl-(Z)-2-(2-phenyl)hydrazineylidene)propanoate2-(2-phenyl)hydrazineylidene)propanoate 2 Ethyl pyruvate (1.6 equiv.) and acetic acid (0.2 equiv.) were added to a solution of the hydrazine (1 equiv., 4 g, 0.022Ethyl mol) pyruvate in EtOH (1.6 (30 mL).equiv.) The and reaction acetic was acid heated (0.2 toequiv.) reflux were for 2 h. added Upon to cooling a solution to room of temperature, the hydrazine the product(1 equiv., crystalized 4 g, 0.022 from mol) the in reactionEtOH (30 mixture. mL). The The reaction crystals were was collectedheated to by reflux filtration for and2 h. washedUpon cooling with cold to roomEtOH. temperature, 2: 5.09 g, 95%, yellowthe product crystals; crystalized1H NMR (400 from MHz, the DMSO- reactiond6) mixture.δ 1.26 (t, J The= 7.1 crys Hz,tals 3H), were 2.05 (s, collected 3H), 4.19 by (q, 13 filtrationJ = 7.1 Hz, and 2H), washed 7.19–7.36 with (m, 3H), cold 9.94 EtOH. (s, 1H). 2: 5.09C NMR g, 95%, (101 yellow MHz, DMSO-d6) crystals;1δH12.4, NMR 14.7, (400 60.8, MHz, 115.6, DMSO- 124.8, 129.4,d6) 133.2, 143.8, 165.2. M.p. 125–127 C. δ 1.26 (t, J = 7.1 Hz, 3H), 2.05 (s,◦ 3H), 4.19 (q, J = 7.1 Hz, 2H), 7.19–7.36 (m, 3H), 9.94 (s, 1H). 13C NMR (101Appendix MHz, A.1.2.DMSO-d6)Synthesis δ 12.4, of ethyl14.7, 160.8,H-indole-2-carboxylate 115.6, 124.8, 129.4, and 133.2, ethyl 143.8, 5-chloro-1H-indole-2-carboxylate 165.2. M.p. 125–127 °C. 3

2 (1 equiv., 6.51 g, 0.027 mol) and polyphosphoric acid (1:10 w/w) were heated at 100 ◦C for 5 h. The reaction Appendixmixture was A.1.2. poured Synthesis onto ice andof ethyl the product 1H-indole-2-carboxylate was extracted with and ethyl ethyl acetate 5-chloro-1H-indole-2- (3 150 mL). The product was × purifiedcarboxylate by silica 3 gel flash column chromatography (crude material: SiO2 = 1:100; ethyl acetate: hexane 1:6 to 1 1:4) to afford 3 (5.8 g, 60%) as a colourless solid. H NMR (400 MHz, DMSO-d6) δ 1.34 (t, J = 7.1 Hz, 3H), 4.34 (q, J =27.1 (1 Hzequiv.,, 2H), 6.51 7.12 g, (dd, 0.027J = 2.2,mol) 0.9 and Hz, polyphosphoric 1H), 7.26 (dd, J = acid8.8, 2.1 (1:10 Hz, w/w 1H),) 7.46were (dt, heatedJ = 8.8, at 0.8 100 Hz, °C 1H), for 7.725 h. 13 (d,TheJ =reaction2.1 Hz, 1H),mixture 12.08 was (s, 1H). pouredC NMR onto (101 ice MHz,and the DMSO- productd6) δ 14.7,was 61.1,extracted 107.6, 114.7,with 121.5,ethyl 125.1,acetate 125.2, (3 × 128.2, 150 129.3,mL). The 136.2, product 161.5. M.p. was 165–166purified◦ C.by silica gel flash column chromatography (crude material: SiO2 = 1:100; ethyl acetate: hexane 1:6 to 1:4) to afford 3 (5.8 g, 60%) as a colourless solid. 1H NMR (400 MHz, DMSO- d6) δ 1.34 (t, J = 7.1 Hz, 3H), 4.34 (q, J = 7.1 Hz, 2H), 7.12 (dd, J = 2.2, 0.9 Hz, 1H), 7.26 (dd, J = 8.8, 2.1 Hz, 1H), 7.46 (dt, J = 8.8, 0.8 Hz, 1H), 7.72 (d, J = 2.1 Hz, 1H), 12.08 (s, 1H). 13C NMR (101 MHz, DMSO-d6) δ 14.7, 61.1, 107.6, 114.7, 121.5, 125.1, 125.2, 128.2, 129.3, 136.2, 161.5. M.p. 165–166 °C.

Appendix A.1.3. Preparation of the Diazonium Salt Solution 4 Br-aniline (1equiv.) was dissolved in a solution of HCl 4M (0.16 w/v) and cooled to 0°C. Then an ice-cold solution of NaNO2 (0.18 w/v) in water was added dropwise and the reaction was stirred for 10min. The diazonium salt was kept at 0°C and used immediately after its preparation.

Appendix A.1.4. Synthesis of Ethyl (E)-3-((4-bromophenyl)diazenyl)-5-chloro-1H-indole-2- carboxylate 4 Compound 3 (1 equiv., 50 mg, 0.22 mmol) and potassium carbonate (10 equiv.) were suspended in DMF (0.2 M) and cooled to 0 °C. Then a solution of the freshly prepared diazonium salt (1.2 equiv.)

Int. J. Mol. Sci. 2020, 21, 334 14 of 20

Appendix A.1.3. Preparation of the Diazonium Salt Solution Int. J. Mol. Sci. 2020, 21, 334 14 of 20 4 Br-aniline (1equiv.) was dissolved in a solution of HCl 4M (0.16 w/v) and cooled to 0◦C. Then an ice-cold solution of NaNO (0.18 w/v) in water was added dropwise and the reaction was stirred for 10min. The diazonium was added dropwise.2 Upon the adding a colourful precipitation formed, the reaction was stirred at 0 salt was kept at 0◦C and used immediately after its preparation. °C for 30 min. The reaction was diluted with water and extracted with ethyl acetate. The product was Appendixpurified by A.1.4. silica Synthesis gel flash of chro Ethylmatography (E)-3-((4-bromophenyl)diazenyl)-5-chloro-1H-indole-2-carboxylate (crude material: SiO2 = 1:100; ethyl acetate: hexane 1:4)4 and yieldedCompound 4 as3 orange(1 equiv., solid 50 mg, (60 0.22 mg, mmol) 70%). and potassium carbonate (10 equiv.) were suspended in DMF (0.2 M)1H and NMR cooled (400 to 0MHz,◦C. Then DMSO- a solutiond6) δ of1.42 the (t, freshly 1H), prepared4.46 (q, J diazonium = 7.2 Hz, salt1H), (1.2 7.43 equiv.) (d, J was= 8.8 added Hz, 1H), dropwise. 7.57 Upon the adding a colourful precipitation formed, the reaction was stirred at 0 C for 30 min. The reaction was (d, J = 8.7 Hz, 1H), 7.78 (d, J = 8.3 Hz, 2H), 7.88 (d, J = 8.3 Hz, 2H), 8.42 (s,◦ 1H), 12.96 (s, 1H). 13C NMR diluted with water and extracted with ethyl acetate. The product was purified by silica gel flash chromatography (crude(101 MHz, material: DMSO- SiO2d=6) 1:100;δ 14.2, ethyl 61.5, acetate: 114.9, hexane117.9, 122.6, 1:4) and 123.5, yielded 123.49,as 126.4, orange 128.4, solid 132.4, (60 mg, 133.7, 70%). 134.3, 152.1, 1 + 160.3.H M.p. NMR 228 (400 °C; MHz, HR-MS DMSO- (ESI):d6) δ 405.996211.42 (t, 1H), [M 4.46 + (q, H]J =; (calcd.7.2 Hz, 1H),405.9952). 7.43 (d, J = 8.8 Hz, 1H), 7.57 (d, J = 8.7 Hz, 13 1H), 7.78 (d, J = 8.3 Hz, 2H), 7.88 (d, J = 8.3 Hz, 2H), 8.42 (s, 1H), 12.96 (s, 1H). C NMR (101 MHz, DMSO-d6) δ 14.2,Appendix 61.5, 114.9, A.1.5. 117.9, Synthesis 122.6, 123.5, of (E)-3-((4-bromophenyl 123.9, 126.4, 128.4, 132.4,)diazenyl)-5-chloro-1H-ind 133.7, 134.3, 152.1, 160.3. M.p.ole-2-carboxylic 228 ◦C; HR-MS Acid (ESI): 405.99621 [M + H]+; (calcd. 405.9952). MTI163 (5) Appendix4 (1 equiv., A.1.5. 50 Synthesis mg, 0.12 of mmol) (E)-3-((4-bromophenyl)diazenyl)-5-chloro-1H-indole-2-carboxylic was suspended in a 2 M solution of NaOH (3 mL) and heated Acid to MTI163 (5) reflux for 3–4h. The reaction was cooled to room temperature acidified with a 2 M solution of HCl and the4 (1 product equiv., 50 was mg, extracted 0.12 mmol) with was ethyl suspended acetate in ato 2 yield M solution 5 (30 mg, of NaOH 60%) (3 as mL) a red and solid. heated 1H to NMR reflux (600 for 3–4 h. The reaction was cooled to room temperature acidified with a 2 M solution of HCl and the product was MHz, DMSO-d6) δ 7.42 (dd, J = 8.7, 2.2 Hz, 1H), 7.55 (d, J =1 8.7 Hz, 1H), 7.78 (d, J = 8.7 Hz, 2H), 7.86 (d, extracted with ethyl acetate to yield 5 (30 mg, 60%) as a red solid. H NMR (600 MHz, DMSO-d6) δ 7.42 (dd, J = 8.7, 2.2J = Hz,8.6 1H),Hz, 7.552H), (d, 8.43J = 8.7(d, Hz,J = 2.1 1H), Hz, 7.78 1H), (d, J =12.868.7 Hz, (s, 2H),1H). 7.86 13C (d,NMRJ = 8.6 (151 Hz, MHz, 2H), 8.43 DMSO- (d, J =d62.1) δ Hz, 114.8, 1H), 118.2, 12.86 13 (s,122.5, 1H). 123.4,C NMR 123.9, (151 126.1, MHz, 128.2, DMSO- 132.4,d6) δ 133.4,114.8, 118.2,133.9, 122.5, 152.0, 123.4, 161.6. 123.9, M.p. 126.1, > 300 128.2, °C, HR-MS 132.4, 133.4, (ESI): 133.9, 377.9636 152.0, + 161.6.[M + H] M.p.+; (calcd.> 300 ◦ C,377.969). HR-MS (ESI): 377.9636 [M + H] ; (calcd. 377.969).

SchemeScheme A2. ReagentsReagents and and conditions conditions (a) ( adiethyl(ethoxymethylene)malonate) diethyl(ethoxymethylene)malonate (1 equiv.), (1 equiv.), toluene, toluene, 110 110°C, ◦24C, h; 24 ( h;b) (bdiphenylether,) diphenylether, 235 235 °C,◦ C,1 1h;h; (c ()c )phosphoryl phosphoryl chloride chloride (1 (1 mL/mmol); mL/mmol); 2 2 h, (d)) p-bromop-bromo phenylhydrazinephenylhydrazine (1.1(1.1 equiv.),equiv.), triethylaminetriethylamine (2.2 (2.2 equiv.), equiv.), ethanol, ethanol, 24 24 h. h.

Appendix A.1.6. Diethyl 2-(((4-chlorophenyl)amino)methylene)malonate 7 Appendix A.1.6. Diethyl 2-(((4-chlorophenyl)amino)methylene)malonate 7 4-Chloroaniline (1 equiv., 8g, 57 mmol) and diethyl(ethoxymethylene)malonate (1 equiv.) were dissolved in toluene4-Chloroaniline (1.25 mL/mmol) (1 and equiv., the reaction 8g, 57 mixture mmol) was and heated diethyl(ethoxymethylene)malonate to reflux. After 22 h the solvent was (1 removed equiv.) underwere reduceddissolved pressure in toluene and the (1.25 residue mL/mmol) was purified and bythe FC reacti (gradienton mixture of 10–30% was EtOAc heated in PEto reflux. and with After a ratio 22 of h 1the/80 1 (weightsolvent (compound)was removed/weight under (silica)) reduced to give pressure 7 (17g., quant.)and the as residue yellow solid.was purifiedH NMR (400by FC MHz, (gradient CDCl3 )ofδ 1.3310– (t, J = 7.1 Hz, 3H), 1.38 (t, J = 7.1 Hz, 3H), 4.25 (q, J = 7.2 Hz, 2H), 4.30 (q, J = 7.2 Hz, 2H), 7.07 (d, J = 8.8 Hz, 2H), 30% EtOAc in PE and with a ratio of 1/80 (weight (compound)/weight13 (silica)) to give 7 (17g., quant.) 7.33 (d, J = 8.8 Hz, 2H), 8.45 (d, J = 13.6 Hz, 1H), 11.00 (br d, J = 13.5 Hz, 1H). C NMR (101 MHz, CDCl3) δ 14.4 (q), 1 14.6as yellow (q), 60.4 solid. (t), 60.7H (t),NMR 94.4 (400 (s), 118.5 MHz, (d), CDCl 130.03 (d),) δ 1.33 130.2 (t, (s), J = 138.1 7.1 Hz, (s), 151.7 3H),(d), 1.38 165.7 (t, J (s),= 7.1 169.1 Hz, (s). 3H), M.p. 4.25 79–81 (q, ◦JC. = 7.2 Hz, 2H), 4.30 (q, J = 7.2 Hz, 2H), 7.07 (d, J = 8.8 Hz, 2H), 7.33 (d, J = 8.8 Hz, 2H), 8.45 (d, J = 13.6 Hz, 1H), 11.00 (br d, J = 13.5 Hz, 1H). 13C NMR (101 MHz, CDCl3) δ 14.4 (q), 14.6 (q), 60.4 (t), 60.7 (t), 94.4 (s), 118.5 (d), 130.0 (d), 130.2 (s), 138.1 (s), 151.7 (d), 165.7 (s), 169.1 (s). M.p. 79–81 °C.

Appendix A.1.7. Synthesis of Ethyl 6-chloro-4-oxo-1,4-dihydroquinoline-3-carboxylate 8 7 (1 equiv., 2g, 7.95 mmol) was dispersed in diphenylether (1.7 mL/mmol), flushed with argon for 5 min and heated to 235 °C for 1 h. The reaction mixture was poured into PE, the formed

Int. J. Mol. Sci. 2020, 21, 334 15 of 20

precipitate was collected by filtration and washed with PE/EtOAc (1/1, 3 × 45 mL) to yield 8 (960 mg, 57%) as a light brown powder.1H NMR (200 MHz, DMSO‐d6) δ 1.28 (t, J = 7.0 Hz, 3H), 4.22 (q, J = 7.1 Hz, 2H), 7.62–7.82 (m, 2H), 8.02–8.12 (m, 1H), 8.59 (d, J = 6.6 Hz, 1H), 12.48 (br s, 1H).13C NMR (151 MHz, DMSO‐d6) δ 14.4 (q), 59.8 (t), 110.1 (s), 121.3 (d), 124.7 (d), 128.4 (s), 129.4 (s), 132.6 (d), 137.7 (s), 145.3 (d), 164.6 (s), 172.3 (s).M.p. > 300 °C.

Int. J.Appendix Mol. Sci. 2020 A.1.8., 21, 334 Synthesis of Ethyl 4,6‐dichloroquinoline‐3‐carboxylate 9 15 of 20

8 (1 equiv., 500 mg, 1.99 mmol) was dispersed in POCl3 (1 mL/mmol) and heated to reflux. After Appendix2 h the A.1.7. reaction Synthesis mixture of was Ethyl poured 6-chloro-4-oxo-1,4-dihydroquinoline-3-carboxylate onto ice, neutralized with satd. NaHCO3, extracted8 with CH2Cl2 (37 (1× 12 equiv., mL/mmol), 2g, 7.95 mmol)washed was with dispersed brine (1 in × diphenylether 12 mL/mmol), (1.7 dried mL/ mmol),over Na flushed2SO4, filtered with argon and for evaporated. 5 min and heatedThe residue to 235 ◦wasC for purified 1 h. The reactionby FC (gradient mixture was of poured5–15% intoEtOAc PE, thein PE formed and precipitate with a ratio was of collected 1/80 (weight by 1 filtration and washed with PE/EtOAc (1/1, 3 45 mL) to yield 8 (9601 mg, 57%) as a light brown powder. H NMR (compound)/weight (silica)) to yield× 9 (406 mg, 76%). H NMR (200 MHz, CDCl3) δ 1.46 (t, J = 7.1 Hz, (200 MHz, DMSO-d6) δ 1.28 (t, J = 7.0 Hz, 3H), 4.22 (q, J = 7.1 Hz, 2H), 7.62–7.82 (m, 2H), 8.02–8.12 (m, 1H), 8.59 (d, 3H), 4.50 (q, J = 7.1 Hz, 2H),13 7.77 (dd, J = 8.9, 2.2 Hz, 1H), 8.08 (d, J = 9.0 Hz, 1H), 8.38 (d, J = 2.2 Hz, J = 6.6 Hz, 1H), 12.48 (br s, 1H). C NMR (151 MHz, DMSO-d6) δ 14.4 (q), 59.8 (t), 110.1 (s), 121.3 (d), 124.7 (d), 13 3 128.41H), (s), 129.4 9.18 (s),(s, 132.61H). (d),C NMR 137.7 (s),(50 145.3 MHz, (d), CDCl 164.6) (s), δ 14.4 172.3 (q), (s). 62.4 M.p. (t),> 300 123.9◦C. (s), 124.5 (d), 127.1 (s), 131.6 (d), 133.0 (d), 134.9 (s), 142.5 (s), 148.0 (s), 150.4 (d), 164.3 (s). M.p. 94–96 °C. Appendix A.1.8. Synthesis of Ethyl 4,6-dichloroquinoline-3-carboxylate 9

Appendix8 (1 equiv., A.1.9. 500 mg, Synthesis 1.99 mmol) of 2 was‐(4‐ dispersedBromophenyl) in POCl‐8‐3chloro(1 mL/‐mmol)2,5‐dihydro and heated‐3H‐pyrazolo[4,3 to reflux. After‐c]quinolin 2 h the ‐ reaction3‐one mixture DCBS192 was poured (10) onto ice, neutralized with satd. NaHCO3, extracted with CH2Cl2 (3 12 mL/mmol), × washed with brine (1 12 mL/mmol), dried over Na2SO4, filtered and evaporated. The residue was purified by FC (gradient9 (1 of equiv., 5–15%× EtOAc 800 mg, in PE 2.98 and mmol) with a ratioand ofp‐ 1bromo/80 (weight phenlhydrazine (compound)/weight hydrochloride (silica)) to yield (1.1 9equiv.)(406 mg, were 1 76%).dispersedH NMR (200in EtOH MHz, (4 CDCl mL/mmol),3) δ 1.46 (t, EtJ 3=N7.1 (2.2 Hz, equiv.) 3H), 4.50 was (q, addedJ = 7.1 Hz,and 2H), the 7.77 reaction (dd, J =mixture8.9, 2.2 Hz,was 1H), heated 13 8.08 (d,to Jreflux= 9.0 Hz, under 1H), 8.38argon (d, Jatmosphere.= 2.2 Hz, 1H), 9.18After (s, 1H).20 h Cthe NMR reaction (50 MHz, mixture CDCl3) δwas14.4 rinsed (q), 62.4 with (t), 123.9 water (s), (2 124.5 (d), 127.1 (s), 131.6 (d), 133.0 (d), 134.9 (s), 142.5 (s), 148.0 (s), 150.4 (d), 164.3 (s). M.p. 94–96 C. mL/mmol), filtered and the precipitate was washed with EtOAc/PE (1/1) (20 mL/mmol).◦ The residue Appendixwas dried A.1.9. under Synthesis reduced of 2-(4-Bromophenyl)-8-chloro-2,5-dihydro-3 pressure to give DCBS192 (1 g, 90%) asH a-pyrazolo[4,3-c]quinolin-3-one yellow solid.1H NMR (400 MHz, DCBS192DMSO (10‐d)6) δ 7.60–7.66 (m, 2H), 7.71–7.76 (m, 2H), 8.16–8.18 (m, 1H), 8.18–8.24 (m, 2H), 8.80 (d, J = 5.5 13 Hz,9 (1 1H), equiv., 12.99 800 (br mg, s, 2.98 1H). mmol) C NMR and p-bromo (151 MHz, phenlhydrazine DMSO‐d6) δ hydrochloride106.1 (s), 116.1 (1.1 (s), equiv.) 119.9 (s), were 120.3 dispersed (d), 121.2 in EtOH(d), (4121.8 mL /(d),mmol), 130.4 Et3 (d),N (2.2 130.8 equiv.) (s), was131.6 added (d), 134.3 and the (s), reaction 139.2 (s), mixture 140.0 was (d), heated 142.4 to (s), reflux 161.6 under (s). M.p. argon >300 atmosphere.°C. HR‐ AfterMS (ESI): 20 h the 373.9694 reaction [M mixture + H]+; was (calcd. rinsed 373.9690). with water (2 mL/mmol), filtered and the precipitate was washed with EtOAc/PE (1/1) (20 mL/mmol). The residue was dried under reduced pressure to give DCBS192 (1 g, 1 90%) as a yellow solid. H NMR (400 MHz, DMSO-d6) δ 7.60–7.66 (m, 2H), 7.71–7.76 (m, 2H), 8.16–8.18 (m, 1H), Appendix A.2. GAUSSIAN Computational Protocol 13 8.18–8.24 (m, 2H), 8.80 (d, J = 5.5 Hz, 1H), 12.99 (br s, 1H). C NMR (151 MHz, DMSO-d6) δ 106.1 (s), 116.1 (s), 119.9 (s), 120.3 (d), 121.2 (d), 121.8 (d), 130.4 (d), 130.8 (s), 131.6 (d), 134.3 (s), 139.2 (s), 140.0 (d), 142.4 (s), 161.6 (s). All calculations were performed+ using the GAUSSIAN 09 software package [43] and the PBE0 M.p. >300 ◦C. HR-MS (ESI): 373.9694 [M + H] ; (calcd. 373.9690). functional, without symmetry constraints. That functional uses a hybrid generalized gradient Appendixapproximation A.2. GAUSSIAN (GGA), Computational including 25% Protocol mixture of Hartree–Fock [44] exchange with DFT [45] exchangeAll calculations‐correlation, were performed given by usingPerdew, the GBurkeaussian and09 Ernzerhof software package functional [43] (PBE) and the [46,47]. PBE0 functional, The basis set withoutused symmetry for the constraints.geometry optimizations That functional uses consisted a hybrid of generalized a standard gradient 6‐31G(d,p) approximation basis set (GGA), for all including atoms [48– 25%54]. mixture The of electronic Hartree–Fock energies [44] exchange obtained with at the DFT PBE0 [45] exchange-correlation, level of theory were given converted by Perdew, to free Burke energy and at Ernzerhof functional (PBE) [46,47]. The basis set used for the geometry optimizations consisted of a standard 6-31G(d,p)298.15 basisK and set 1 for atm all by atoms using [48 zero–54]. point The electronic energy and energies thermal obtained energy at thecorrections PBE0 level based of theory on structural were convertedand vibration to free energy frequency at 298.15 data K and calculated 1 atm by at using the zerosame point level. energy and thermal energy corrections based on structural and vibration frequency data calculated at the same level.

FigureFigure A1. GaussianA1. Gaussian calculation calculation of two configurationsof two configurations of MTI163. of The MTI163. structure The forming structure a 7-membered forming a 7‐ a ringmembered via an intramolecular ring via an intramolecular hydrogen bond hydrogen ( ) is approximately bond (a) is approximately 11 kcal/mol more11 kcal/mol stable more than stable a conformation without such an intramolecular H-bond (b). than a conformation without such an intramolecular H‐bond (b).

Int. J. Mol. Sci. 2020, 21, 334 16 of 20 Int. J. Mol. Sci. 2020, 21, 334 16 of 20 Appendix A.3. Supplementary Figures from Computational Docking Int. J. Mol. Sci. 2020, 21, 334 16 of 20 Appendix A.3. Supplementary Figures from Computational Docking Appendix A.3. Supplementary Figures from Computational Docking

Figure A2. Diversity of etomidate binding poses in 6HUP. Of the top 100 poses (GoldScore ranking), consensus scoring and ligand flexibility suggests that the diversity of the posing space is very high FigureandFigure experimentally A2. A2. DiversityDiversity guidedof of etomidate etomidate constraints binding binding or poses a poses higher in in6HUP. 6HUP.level Ofof Of thecomputational the top top 100 100 poses poses theory (GoldScore (GoldScore is needed. ranking), ranking), The consensusdiversityconsensus of scoring scoringthe posing and and spaceligand ligand is flexibility flexibilityillustrated suggestssuggests based on thatthat the the themost diversity diversity diverse of ofsolutions the the posing posing among space space the is veryis top very 20 high fromhigh and andtwoexperimentally experimentallyscoring functions. guided guided constraintsAll posesconstraints display or a higher or a highhigher level overlap of level computational ofof computational ligand–protein theory is theory needed.interactions is The needed. diversitywith Thethe of diversitydiazepamthe posing of present the space posing isin illustratedthis space site isin illustrated based6HUP. on (a the) based Front most onview; diverse the (mostb) solutionsside diverse view. among solutions the among top 20 fromthe top two 20 scoring from twofunctions. scoring All functions. poses display All poses a high display overlap a of high ligand–protein overlap of interactions ligand–protein with theinteractions diazepam with present the in a b diazepamthis site inpresent 6HUP. in ( )this Front site view; in 6HUP. ( ) side (a) Front view. view; (b) side view.

Figure A3. Different binding poses of MTI163 in 6HUP. Top 20 poses from GoldScore and ChemScore Figure A3. Different binding poses of MTI163 in 6HUP. Top 20 poses from GoldScore and ChemScore ranking were analysed, and the seven poses thatremain among top 20 by consensus (i.e., in ranks 1–20 ranking were analysed, and the seven poses thatremain among top 20 by consensus (i.e., in ranks 1– in both scoring functions) are depicted here. (a) Front view; (b) Side view. Figure20 in both A3. scoringDifferent functions) binding posesare depicted of MTI163 here. in (6HUP.a) Front Top view; 20 poses(b) Side from view. GoldScore and ChemScore ranking were analysed, and the seven poses thatremain among top 20 by consensus (i.e., in ranks 1– 20 in both scoring functions) are depicted here. (a) Front view; (b) Side view.

Int. J. Mol. Sci. 2020, 21, 334 17 of 20 Int.Int. J.J. Mol.Mol. Sci.Sci. 2020,, 2121,, 334334 17 of 20

Figure A4. A4. DifferentDifferent binding binding poses of MTI163 MTI163 in in the the N265 N265SS mutant. mutant. There There is is no no ligand ligand bound bound structure structure with the β1 subunit, thus 6HUP was point mutated for the docking. In In this case, five five docking poses were found in the top 20 of GoldScore and ChemScore. ( a)) Front Front view; view; ( (b)) Side Side view; view; ( (c)) Comparison Comparison between the the pose pose displayed displayed in in Figure Figure 5 5and and the the most most similarsimilar similar posepose pose observedobserved observed inin thethe in theN265SN265S N265S mutant.mutant. mutant. TheThe twoThetwo twoboundbound bound complexescomplexes complexes areare aresuperposedsuperposed superposed (reflected(reflected (reflected bybyby thethe the two-coloredtwo-colored two-colored ribbonribbon ribbon rendering),rendering), rendering), thethe the 6HUP6HUP (wild(wild typetypetypeβ β3)3) pose pose and and N265 N265 are are rendered rendered as light as light green gr sticks,een sticks, the S265 the and S265 the and corresponding the corresponding MTI163 MTI163pose are pose rendered are rendered as yellow as sticks. yellow sticks. Appendix A.4. MTI163 Dose Response Curve in α1β1Receptors Appendix A.4. MTI163 Dose Response Curve in α1β1Receptors

α β Figure A5. Concentration-dependent MTI1 MTI16363 modulation modulation of of GABA EC 3–53–5 elicitedelicitedelicited currentcurrent current inin in α11β1.1. Response at each point of measurement is represented as individual data point as Mean SEM (n = 4, Response at each point of measurement is represented as individual data point as Mean± ± SEM (n = EC = 7.65 µM, nH = 0.78). 4, EC50 5050 == 7.657.65 µM,µM, nHnH == 0.78).0.78).

References

1. Sieghart, W.; Savi´c,M.M. International Union of Basic and Clinical Pharmacology. CVI: GABAA Receptor Subtype- and Function-selective Ligands: Key Issues in Translation to Humans. Pharmacol. Rev. 2018, 70, 836–878. [CrossRef][PubMed] 2. Jembrek, M.J.; Vlainic, J. GABA Receptors: Pharmacological Potential and Pitfalls. Curr. Pharm. Des. 2015, 21, 4943–4959. [CrossRef][PubMed]

3. Sigel, E.; Buhr, A. The benzodiazepine binding site of GABAA receptors. Trends Pharmacol. Sci. 1997, 18, 425–429. [CrossRef]

Int. J. Mol. Sci. 2020, 21, 334 18 of 20

4. Masiulis, S.; Desai, R.; Ucha´nski,T.; Martin, I.S.; Laverty, D.; Karia, D.; Malinauskas, T.; Zivanov, J.; Pardon, E.;

Kotecha, A.; et al. GABAA receptor signalling mechanisms revealed by structural pharmacology. Nature 2019, 565, 454–459. [CrossRef][PubMed]

5. Walters, R.J.; Hadley, S.H.; Morris, K.D.; Amin, J. Benzodiazepines act on GABAA receptors via two distinct and separable mechanisms. Nat. Neurosci. 2000, 3, 1274–1281. [CrossRef]

6. Maldifassi, M.C.; Baur, R.; Sigel, E. Functional sites involved in modulation of the GABAA receptor channel by the intravenous anesthetics propofol, etomidate and pentobarbital. Neuropharmacology 2016, 105, 207–214. [CrossRef] 7. Chiara, D.C.; Dostalova, Z.; Jayakar, S.S.; Zhou, X.; Miller, K.W.; Cohen, J.B. Mapping general anesthetic binding site(s) in human α1β3 γ-aminobutyric acid type A receptors with [3H]TDBzl-etomidate, a photoreactive etomidate analogue. Biochemistry 2012, 51, 836–847. [CrossRef]

8. Olsen, R.W. GABAA receptor: Positive and negative allosteric modulators. Neuropharmacology 2018, 136, 10–22. [CrossRef] 9. Solomon, V.R.; Tallapragada, V.J.; Chebib, M.; Johnston, G.; Hanrahan, J.R. GABA allosteric modulators: An overview of recent developments in non-benzodiazepine modulators. Eur. J. Med. Chem. 2019, 171, 434–461. [CrossRef]

10. Sieghart, W. Chapter Three—Allosteric Modulation of GABAA Receptors via Multiple Drug-Binding Sites. In Advances in Pharmacology; Rudolph, U., Ed.; Academic Press: Cambridge, MA, USA, 2015; pp. 53–96. 11. Puthenkalam, R.; Hieckel, M.; Simeone, X.; Suwattanasophon, C.; Feldbauer, R.V.; Ecker, G.F.; Ernst, M.

Structural Studies of GABAA Receptor Binding Sites: Which Experimental Structure Tells us What? Front. Mol. Neurosci. 2016, 9, 44. [CrossRef] 12. Iorio, M.T.; Rehman, S.; Bampali, K.; Stoeger, B.; Schnürch, M.; Ernst, M.; Mihovilovic, M.D. Variations on a

scaffold—Novel GABAA receptor modulators. Eur. J. Med. Chem. 2019, 180, 340–349. [CrossRef][PubMed] 13. Ramerstorfer, J.; Furtmüller, R.; Sarto-Jackson, I.; Varagic, Z.; Sieghart, W.; Ernst, M. The GABAA Receptor α+β Interface: A Novel Target for Subtype Selective Drugs. J. Neurosci. 2011, 31, 870–877. [CrossRef] − [PubMed]

14. Maldifassi, M.C.; Baur, R.; Sigel, E. Molecular mode of action of CGS 9895 at α1β2γ2 GABAA receptors. J. Neurochem. 2016, 138, 722–730. [CrossRef][PubMed] 15. Varagic, Z.; Ramerstorfer, J.; Huang, S.; Rallapalli, S.; Sarto-Jackson, I.; Cook, J.; Sieghart, W.; Ernst, M. Subtype selectivity of α+β site ligands of GABA receptors: Identification of the first highly specific − A positive modulators at α6β2/3γ2 receptors. Br. J Pharm. 2013, 169, 384–399. [CrossRef][PubMed] 16. Siebert, D.C.B.; Bampali, K.; Puthenkalam, R.; Varagic, Z.; Sarto-Jackson, I.; Scholze, P.; Sieghart, W.; Mihovilovic, M.D.; Schnürch, M.; Ernst, M. Engineered Flumazenil Recognition Site Provides Mechanistic

Insight Governing Benzodiazepine Modulation in GABAA Receptors. ACS Chem. Boil. 2018, 13, 2040–2047. [CrossRef]

17. Zhu, S.; Noviello, C.M.; Teng, J.; Walsh, R.M.; Kim, J.J.; Hibbs, R.E. Structure of a human synaptic GABAA receptor. Nature 2018, 559, 67–72. [CrossRef] 18. Simeone, X.; Siebert, D.C.; Bampali, K.; Varagic, Z.; Treven, M.; Rehman, S.; Pyszkowski, J.; Holzinger, R.;

Steudle, F.; Scholze, P.; et al. Molecular tools for GABAA receptors: High affinity ligands for β1-containing subtypes. Sci. Rep. 2017, 7, 5674–5685. [CrossRef] 19. Thomet, U.; Baur, R.; Scholze, P.; Sieghart, W.; Sigel, E. Dual mode of stimulation by the beta-carboline ZK

91085 of recombinant GABAA receptor currents: Molecular determinants affecting its action. Br. J. Pharmacol. 1999, 127, 1231–1239. [CrossRef] 20. Wingrove, P.B.; Wafford, K.A.; Bain, C.; Whiting, P.J. The modulatory action of loreclezole at the gamma-aminobutyric acid type A receptor is determined by a single amino acid in the beta 2 and beta 3 subunit. Proc. Natl. Acad. Sci. USA 1994, 91, 4569–4573. [CrossRef] 21. Desai, R.; Ruesch, D.; Forman, S.A. Gamma-amino butyric acid type A receptor mutations at beta2N265 alter etomidate efficacy while preserving basal and agonist-dependent activity. Anesthesiology 2009, 111, 774–784. [CrossRef] 22. Wafford, K.; Bain, C.; Quirk, K.; McKernan, R.; Wingrove, P.; Whiting, P.; Kemp, J. A novel allosteric

modulatory site on the GABAA receptor β subunit. Neuron 1994, 12, 775–782. [CrossRef] Int. J. Mol. Sci. 2020, 21, 334 19 of 20

23. Thompson, S.A.; Wingrove, P.B.; Connelly, L.; Whiting, P.J.; Wafford, K.A. Tracazolate Reveals a Novel Type

of Allosteric Interaction with Recombinant γ-Aminobutyric AcidA Receptors. Mol. Pharmacol. 2002, 61, 861–869. [CrossRef][PubMed] 24. Stevenson, A.; Wingrove, P.B.; Whiting, P.J.; Wafford, K.A. beta-Carboline gamma-aminobutyric acidA receptor inverse agonists modulate gamma-aminobutyric acid via the loreclezole binding site as well as the benzodiazepine site. Mol. Pharmacol. 1995, 48, 965–969. [PubMed] 25. Fernandez, S.P.; Karim, N.; Mewett, K.N.; Chebib, M.; Johnston, G.A.; Hanrahan, J.R. Flavan-3-ol esters:

New agents for exploring modulatory sites on GABAA receptors. Br. J. Pharmacol. 2012, 165, 965–977. [CrossRef][PubMed] 26. Belelli, D.; Lambert, J.J.; Peters, J.A.; Wafford, K.; Whiting, P.J. The interaction of the general anesthetic etomidate with the gamma-aminobutyric acid type A receptor is influenced by a single amino acid. Proc. Natl. Acad. Sci. USA 1997, 94, 11031–11036. [CrossRef] 27. Hanrahan, J.R.; Chebib, M.; Johnston, G.A.R. Chapter Seven—Interactions of Flavonoids with Ionotropic GABA Receptors. In Advances in Pharmacology; Rudolph, U., Ed.; Academic Press: Cambridge, MA, USA, 2015; pp. 189–200. 28. Kavvadias, D.; Monschein, V.; Sand, P.; Riederer, P.; Schreier, P. Constituents of Sage (Salvia officinalis) with in vitro Affinity to Human Brain Benzodiazepine Receptor. Planta Med. 2003, 69, 113–117. [CrossRef] 29. Kavvadias, D.; Sand, P.; A Youdim, K.; Qaiser, M.Z.; Rice-Evans, C.; Baur, R.; Sigel, E.; Rausch, W.-D.; Riederer, P.; Schreier, P. The flavone hispidulin, a benzodiazepine receptor ligand with positive allosteric properties, traverses the blood–brain barrier and exhibits anticonvulsive effects. Br. J. Pharmacol. 2004, 142, 811–820. [CrossRef] 30. Sigel, E.; Baur, R.; Furtmueller, R.; Razet, R.; Dodd, R.H.; Sieghart, W. Differential Cross Talk of ROD Compounds with the Benzodiazepine Binding Site. Mol. Pharmacol. 2001, 59, 1470–1477. [CrossRef] 31. El Hadri, A.; Abouabdellah, A.; Thomet, U.; Baur, R.; Furtmüller, R.; Sigel, E.; Sieghart, W.; Dodd, R.H.

N-Substituted 4-amino-3,3-dipropyl-2(3H)-furanones: New positive allosteric modulators of the GABAA receptor sharing electrophysiological properties with the anticonvulsant loreclezole. J. Med. Chem. 2002, 45, 2824–2831. [CrossRef] 32. Stornaiuolo, M.; De Kloe, G.E.; Rucktooa, P.; Fish, A.; Van Elk, R.; Edink, E.S.; Bertrand, D.; Smit, A.B.; De Esch, I.J.P.; Sixma, T.K. Assembly of a π–π stack of ligands in the binding site of an acetylcholine-binding protein. Nat. Commun. 2013, 4, 1875. [CrossRef] 33. Richter, L.; De Graaf, C.; Sieghart, W.; Varagic, Z.; Mörzinger, M.; De Esch, I.J.P.; Ecker, G.F.; Ernst, M.

Diazepam-bound GABAA receptor models identify new benzodiazepine binding-site ligands. Nat. Methods 2012, 8, 455–464. [CrossRef][PubMed] 34. Thomet, U.; Baur, R.; Razet, R.; Dodd, R.H.; Furtmüller, R.; Sieghart, W.; Sigel, E. A novel positive allosteric

modulator of the GABAA receptor: The action of (+)-ROD188. Br. J. Pharmacol. 2000, 131, 843–850. [CrossRef] [PubMed] 35. Armarego, W.L.F.; Lin Chai, C.L. Purification of Laboratory Chemicals, 7th ed.; Butterworth-Heinemann: Oxford, UK, 2012. 36. Varagic, Z.; Wimmer, L.; Schnürch, M.; Mihovilovic, M.D.; Huang, S.; Rallapalli, S.; Cook, J.M.; Mirheydari, P.; Ecker, G.F.; Sieghart, W.; et al. Identification of novel positive allosteric modulators and null modulators at the GABA receptor α+β interface. Br. J. Pharmacol. 2013, 169, 371–383. [CrossRef] A − 37. Minier, F.; Sigel, E. Positioning of the α-subunit isoforms confers a functional signature to γ-aminobutyric acid type A receptors. Proc. Natl. Acad. Sci. USA 2004, 101, 7769–7774. [CrossRef][PubMed] 38. Baumann, S.W.; Baur, R.; Sigel, E. Subunit Arrangement of γ-Aminobutyric Acid Type A Receptors. J. Biol. Chem. 2001, 276, 36275–36280. [CrossRef][PubMed]

39. Baumann, S.W.; Baur, R.; Sigel, E. Forced Subunit Assembly in α1β2γ2 GABAAReceptors: INSIGHT INTO THE ABSOLUTE ARRANGEMENT. J. Biol. Chem. 2002, 277, 46020–46025. [CrossRef][PubMed] 40. Sieghart, W.; Schuster, A. Affinity of various ligands for benzodiazepine receptors in rat cerebellum and hippocampus. Biochem. Pharmacol. 1984, 33, 4033–4038. [CrossRef] 41. Yung-Chi, C.; Prusoff, W.H. Relationship between the inhibition constant (KI) and the concentration of inhibitor which causes 50 per cent inhibition (I50) of an enzymatic reaction. Biochem. Pharmacol. 1973, 22, 3099–3108. [CrossRef] Int. J. Mol. Sci. 2020, 21, 334 20 of 20

42. Jones, G.; Willett, P.; Glen, R.C.; Leach, A.R.; Taylor, R. Development and validation of a genetic algorithm for flexible docking11Edited by F. E. Cohen. J. Mol. Biol. 1997, 267, 727–748. [CrossRef] 43. Gaussian, R.A.; Frisch, M.J.; Trucks, G.W.; Schlegel, H.B.; Scuseria, G.E.; Robb, M.A.; Cheeseman, J.R.; Scalmani, G.; Barone, V.; Mennucci, B.; et al. Gaussian, Inc., Wallingford, CT, USA. 2009. Available online: http://wild.life.nctu.edu.tw/~{}jsyu/compchem/g09/g09ur/m_citation.htm (accessed on 3 January 2020). 44. Hehre, W.J. Ab initio molecular orbital theory. Accounts Chem. Res. 1976, 9, 399–406. [CrossRef] 45. Parr, R.G.Y. Density Functional Theory of Atoms and Molecules; Oxford University Press: New York, NY, USA, 1989. 46. Perdew, J.P. Density-functional approximation for the correlation energy of the inhomogeneous electron gas. Phys. Rev. B 1986, 33, 8822–8824. [CrossRef][PubMed] 47. Perdew, J.P.; Burke, K.; Ernzerhof, M. Generalized Gradient Approximation Made Simple. Phys. Rev. Lett. 1996, 78, 1396. [CrossRef] 48. Binning, R.C., Jr.; Curtiss, L.A. Compact contracted basis sets for third-row atoms: Ga–Kr. J. Chem. Physics. 1990, 11, 1206–1216. [CrossRef] 49. Hay, P.J. Gaussian basis sets for molecular calculations. The representation of 3d orbitals in transition-metal atoms. J. Chem. Phys. 1977, 66, 4377–4384. [CrossRef] 50. Krishnan, R.; Binkley, J.S.; Seeger, R.; Pople, J.A. Self-consistent molecular orbital methods. XX. A basis set for correlated wave functions. J. Chem. Phys. 1980, 72, 650–654. [CrossRef] 51. McGrath, M.P.; Radom, L. Extension of Gaussian-1 (G1) theory to bromine-containing molecules. J. Chem. Phys. 1991, 94, 511–516. [CrossRef] 52. McLean, A.D.; Chandler, G.S. Contracted Gaussian basis sets for molecular calculations. I. Second row atoms, Z = 11–18. J. Chem. Phys. 1980, 72, 5639–5648. [CrossRef] 53. Raghavachari, K.; Trucks, G.W. Highly correlated systems. Excitation energies of first row transition metals Sc-Cu. J. Chem. Phys. 1989, 91, 1062–1065. [CrossRef] 54. Wachters, A.J.H. Gaussian Basis Set for Molecular Wavefunctions Containing Third-Row Atoms. J. Chem. Phys. 1970, 52, 1033–1036. [CrossRef]

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).