Structure-based Discovery of Novel Ligands for the Orexin 2 Receptor

The orexin receptors are peptide-sensing G protein-coupled receptors that are intimately linked with regulation of the sleep/wake cycle. We used a recently solved X-ray structure of the orexin receptor subtype 2 in computational docking calculations with the aim to identify additional ligands with unprecedented chemotypes. We found validated ligands with a high hit rate of 29% out of those tested, none of them showing selectivity with respect to the orexin receptor subtype 1. Furthermore, of the higher-affinity compounds examined, none showed any agonist activity. While novel chemical structures can thus be found, selectivity is a challenge owing to the largely identical binding pockets.


Introduction
The sleep/wake cycle is one of the fundamental features of organisms with a central nervous system.Especially in humans, it guides our every-day lives and is one of the key activities that keep us healthy and sane.Both sleep deprivation and the inability to sleep (insomnia) are therefore unpleasant for an individual and have been shown to have harmful physiological effects (1), and constitute a huge burden on society as a whole.(2) In order to ameliorate insomnia, two ways can be envisioned: one is to make the brain artificially sleepy; the other is to block the signals that mediate wakefulness.While more traditional insomnia medications have tried to do the former, the latter seems like a strategy that should lead to fewer side effects.
In humans, one of the main pathways to transmit wakefulness signals is the orexin peptide/orexin receptor system.(3) It consists of two peptides, orexin A and orexin B, and two G protein-coupled receptors (GPCRs), orexin receptor subtype 1 (OX1R) and subtype 2 (OX2R).The two peptides are 33 and 28 amino acids in length, respectively, and bind with nanomolar affinities to both receptors.At the same time, the two receptors are highly homologous, with 63% sequence identity, and, most importantly, differ by only two amino acid substitutions in the binding pocket.Despite this high sequence similarity, orexin A achieves a certain amount of selectivity for OX1R over OX2R.(4) In 2014, the first antagonist of the OX2R, suvorexant, obtained regulatory approval as an insomnia medication in the US.One year later, the crystal structure of this molecule in complex with OX2R was published.(5) It showed that suvorexant was binding in a "horseshoe shape", which is consistent with earlier NMR investigations showing that suvorexant was able to adopt such a conformation in solution.(6) The receptor was confirmed to be a typical GPCR with seven membrane-spanning helices.Moreover, the crystallographic data demonstrated that suvorexant was interacting with the receptor mainly through hydrophobic interactions.In fact, there are only two polar interactions, one a direct hydrogen bond to Asn324 6.55 , the other a water-mediated hydrogen bond with His350 7.39 (Figure 1).(5) Aiming to exploit this structure in an unbiased computational screen, we have asked three questions.Firstly, could an unbiased docking screen and the analysis of the predicted binding modes suggest other polar interaction possibilities for small molecules in the binding pocket?Secondly, could an apolar binding pocket as this one be as suitable for docking (manifested by high hit rates) as the more polar pockets of the aminergic receptors investigated earlier.(7,8,9,10,11) Lastly, could comparison of our hit molecules with the already known chemical space for the orexin receptors identify novel chemotypes that might present scaffolds worthwhile to elaborate on further with medicinal chemistry?Figure 1: Two-dimensional representation of the binding mode of suvorexant (PDB code 4S0V) generated with the MOE suite.Two interactions of suvorexant with OX2R are indicated as dashed green lines: An H-bond between amide carbonyl and Asn324 6.55 and a π-π interaction between 5-methyl-2-triazolphenyl and His350 7.39 .

Selection of parent hits: Primary docking screen
Docking of 7.337 million compounds from the DrugsNow subset of the ZINC database into the OX2R (see Methods for details) and subsequent post-processing of its outcome resulted in 6500 unique poses, split up in 6000 from the top ranks of the lists sorted according to HYBRID, DSX, or SZYBKI scores and an additional 500 from the calculations with DOCK.The visual inspection of these poses resulted in 85 virtual hit candidates (HYBRID: 14; DSX: 25; SZYBKI: 14 and DOCK: 32 poses, respectively [Figure 2]).Forty-three compounds were available from their vendors and analyzed pharmacologically.These compounds will be referred to as the "parent" compounds P1-P43 in the following and are shown in Table S2, with selected compounds also displayed in Figure 4.

Validation of cell lines and examination of primary docking screen
Pharmacological studies were conducted in CHO cells stably expressing the human orexin 1 receptor (CHO-OX1) or human orexin 2 receptor (CHO-OX2).Suvorexant, a known dual OX1/OX2 receptor antagonist, inhibited the specific binding of both 3 H-SB674042 to the human OX1R and  Selected virtual hits were validated experimentally.found to have much higher affinity for the CHO-OX1 cells than the CHO-OX2 cells, whereas the known OX2R antagonists JNJ10397049 and EMPA had much higher affinity for the CHO-OX2 cells (Table 1).These are consistent with previous values and demonstrate the existence of each of the receptors in the two respective cell lines.(12,13,14) Conversely, in cells not transfected with either receptor, no binding of 3 H-SB674042 or 3 H-EMPA, respectively, could be observed.
Of the 43 initial compounds identified from the primary docking screen, eleven had some degree of measurable affinity for either the OX1 or OX2 receptor, giving an initial hit rate of 26 %.Of these, P33 had the highest affinity with a pKi value of 5.54 at the OX2R (Figure 3, Table 1, Figure 4, for space reasons, all other compounds that are not among the ten with the highest affinity are depicted in the Supplementary Information [Table S 2]).Almost all measured compounds did not show selectivity.Five compounds showed a very minor degree of selectivity: P34 had a 5-fold OX2R selectivity (ΔpKi of -0.69), P9 and P33 a 4-fold OX2R selectivity (a ΔpKi of -0.62 and -0.59, respectively) whilst P12 and P18 had at least a 3-fold OX1R selectivity.For nine parent compounds, daughter derivatives (F) were obtained: We selected between one (P22 and P31) and 15 (P35) daughter molecules from each parent -in total 54 compounds -based on availability of similar compounds and their favorable docking scores and poses.The identifiers of the daughter compounds in this text are based on the ID of the parent compound and the order number of each of them, separated by a point.E.g. compound F33.1 is the first daughter compound of parent P33.Table 2 shows their affinities measured against OX1R and OX2R.Compound P33 can be considered the most fruitful parent compound, as all but one of the seven derivatives measured also showed measurable affinity against the OX2R.Moreover, its daughter F33.3 bound with the highest affinity of all ligands, at a pKi of 6.18 (Ki of 660 nM, Figure 3).In total, 16 of the 54 derivatives tested bound to the receptor.This brings the hit rate of daughter compounds to 30 %. Finally, to further exclude any non-orexin receptor mediated effects, the most potent ligands were examined for their ability to bind to an unrelated receptor, the human β1adrenergic receptor.Although the β1antagonist CGP20712A inhibited specific binding with high affinity, none of the ligands with orexin receptor affinity had any detectable binding for the β1-adrenergic receptor (see supplementary data).

Examination of functional responses
Orexin 1 and 2 receptors are both Gqcoupled GPCRs and therefore stimulate an increase in intracellular calcium release.To examine whether the compounds had an agonist activity, the ability of some of the higher affinity compounds to stimulate intracellular calcium release was measured.Orexin A stimulated a potent agonist response in both the CHO-OX1 cells (pEC50 9.67 ± 0.13, 19.4 ± 24 fold over basal, 58.8 ± 2.4 % that of 10 μM ionomycin, n=5) and the CHO-OX2 cells (pEC50 10.31 ± 0.07, 32.8 ± 4.2 fold over

Structure-Activity Relationship
The lack of polar features in the binding pocket is echoed in the ligands.Therefore, it is challenging to develop a clear-cut SAR for them.There are many chemical solutions for a molecule to form favorable interactions with this receptor.In particular, apolar interactions are relatively tolerant towards subtle geometric changes.This tolerance can be seen here as several bulky substituents had very little effect of affinity.
Looking at P33 and its derivatives (F33.1,F33.2, F33.3, [F33.4:no affinity], F33.5, F33.6, and F33.7), all these points are confirmed.This mini-series consists of the most potent parent P33 and daughter molecule (F33.3) with the highest number of active daughter compounds (six).The only polar interaction these molecules form is with Asn324 6.55 , yet the overall binding mode of P33 is not suvorexant-like (Figure 6 right).The daughters tell us that the position or existence of the halogens on the benzene rings are not important (F33.2 and F33.3 bind), but that more bulk in this region is not favorable (F33.1 is worse).There seems to be more space around the methyl-furane, however, as it can be replaced with bulkier groups (F33.6,F33.7).For these latter molecules, a more suvorexantlike binding mode is conceivable.Interestingly, all the most active compounds show a double acceptor feature, which is absent in the weaker ligands.The most potent compound of this series, F33.3, might also be the only one truly capable of adapting a suvorexant-like conformation according to docking (Figure 6 left).
For P7 and similar (F7.1, [F7.2: no affinity], F7.3, F7.4, [F7.5 and F7.6: no affinity], and F7.7), a slightly different picture emerges.Compound P7 interacts with residues Glu212 45.52 and Arg328 6.59 , both located at the entrance of the binding site and forming a lid, shielding it from bulk solvent.The derivatives are only bound to Asn324 6.55 .Of note, F7.7 binds in an orientation reminiscent of a reverse suvorexant binding mode (Figure 7).Despite a certain difference in the overall layout, all of the compounds with affinity feature a pyrimidinone ring and a relatively apolar double ring system.
Finally, as a negative example, we turn to parent compound P9.Although P9 has affinity for orexin receptors, none of the 12 derivatives showed much binding.This series also shows the perils of similarity searches, as several of the ligands would not be considered similar by a chemist.Yet, these seemed to interact favorably with the binding pocket in docking.F9.8 is the only compound with a similar "double lactam" cyclohexene ring and the only to show some effect.Overall, our docking screen showed that even for relatively featureless apolar binding pockets, ligands can be found in large databases based on shape complementarity (see Figure 4 for the ten most affine compounds found in this study).Interestingly, the scarcity of strong polar or even chargecharge interaction possibilities did not hamper docking's ability to find ligands, evident from the overall hit rate of 29%, which compares favorably with other studies on class A GPCRs.Several novel chemotypes were identified and might serve as seeds for further development.

Discussion and conclusions
Our unbiased docking screen (using the OX2R crystal structure) answers the three questions posed in the introduction.A metric that is often used as the figure of merit in a docking screen is the hit rate, defined as the percentage of all tested molecules that are found to interact with the receptor.In this study, taking the hits from the parent and daughter screen into account, a hit rate of 29% was achieved in the CHO-OX2 cells, based on binding affinity.Despite the challenging nature of the binding pocket, with only few polar interactions sprinkled throughout, this puts the present screen in the same league as previous studies with polar or even charged features within the binding pockets.Docking itself thus does not seem to be hampered by this comparative scarcity of directional interactions in the OX2R.The hit rates of the daughter generation are almost the same as the parents' rates, however.This might be because the similarity calculations were based on global similarities.A stricter similarity search, retaining the scaffolds of the parent compounds, might have produced higher hit rates.
Secondly, we took a closer look at the predicted binding modes and compared them with the ones of suvorexant and SB-674042.The majority of molecules interact with very similar residues, predominantly Asn324 6.55 and His350 7.39 .Interestingly, however, a few of them also formed interactions with Thr111 2.61 , a residue that differs in OX1R, where it is a serine (Ser103 2.61 ).Despite this being a small deviation, such compounds might be starting points for selective orexin receptor antagonists (SORAs).On top of this, compound P22 forms a hydrogen bond with the backbone carbonyl of Pro131 3.29 , which has so far not been observed in any of the crystal structures.
Lastly, this docking strategy has managed to identify several scaffolds that can be considered novel by common chemoinformatic criteria.Among them are compound P27, whose ECFP4 Tanimoto similarity to its closest match amongst any known ligands of OX2R is only 0.237 (median similarity of 0.115).Two additional compounds, F33.2 and F7.4,show also distant closest matches and low maximum similarities of 0.241 (median of 0.124) and 0.254 (median of 0.121), respectively.In total, 52 compounds that show some affinity in our assay, had an ECFP4 Tanimoto similarity to their closest neighbors of 0.45 or less, and 22 with a similarity value of 0.30 and less, commonly regarded as a threshold indicating dissimilarity (ECFP4 Tanimoto similarity values and ROCS TanimotoCombo scores can be found in the Supplementary Information [Table S

and Table S 5]).
A key question for future investigations is whether these scaffolds can easily be derivatized.To answer this question, the composition was analyzed with the PINGUI (15) toolbox, which we have developed earlier.Each of the compounds shows a facile synthetic breaking point, yielding fragments of a size that is frequently occurring in building block databases.The number of applicable reactions to fragments of our hits after their retrosynthetical decomposition ranges from two (F21.3) to nine (P18) (Table S 6).It is likely that also the derivatives will show affinity, as we have already exhausted existing chemical space through SAR-by-catalog.In particular compound P33 seems very promising, as five of its daughter compounds (F33.1-3,F33.6 and F33.7) also show binding to the receptor.Our analysis found 103 retrosynthetical disconnections.Four unique reactions can be applied to the resulting fragment set.It has to be noted that P33 shares some moieties with the known ligands EMPA and SB-649868.Based on our SAR considerations and our experience from similar projects (15,16) we are convinced that there is room for diversity around the common rings, however.
By way of better characterization, we also tested our compounds against the OX1R, despite the fact that we never made a prediction about selectivity.Not unexpectedly, most of the compounds behave as DORAs.In fact, our OX2R-focused docking produced five compounds with a measurable affinity selectivity between OX2R and OX1R, but only up to a maximum selectivity of 5-fold (a ΔpKi of 0.6).Although measurable, these values are low and at the detection limit of the assay.At the same time, OX2R-selective compounds are generally regarded as preferable for the treatment of insomnia, as they carry lower risks of OX1R-mediated side effects, such as compulsive behavior and substance misuse disorders.(17) Would this be different if we had docked to both subtypes?In order to answer this question, we did a dual docking as described before.(11) Briefly, the same small molecule dataset was docked to both the OX1R and the OX2R.For the OX1R, the X-ray structure with PDB ID 4ZJ8 ( 18) was used and prepared in an identical fashion as its OX2R counterpart.After docking, molecules were reranked in order to favor molecules that would display the desired selectivity.(11) The molecules populating the top 500 ranks of this docking were all distinct from the ones of the original docking against the OX2R, and, therefore, there was also no overlap between the sets of assayed molecules.However, even with this differential docking, we did not find ligands with more than a ΔpKi of 0.5.The data for these molecules is presented in Table 3.It thus stands to reason that this featureless binding pocket presents a case where standard protocols run in high-throughput mode do not discriminate well enough in order to predict selectivity between the two orexin receptor subtypes.As we have shown recently, this is different for dedicated more precise docking calculations used to optimize ligand selectivity, driving it towards higher OX1R affinity.(19) In summary, our in silico screen yielded several compounds with potential for further development as orexin receptor ligands and shed light on possible alternative interaction patterns that can be exploited in future screening and optimization work.

Receptor preparation
The crystal structure of the OX2R liganded by suvorexant (PDB code 4S0V) (5) was prepared using the MOE software suite.(20) Briefly, i) all water molecules, metal ions, precipitants and lipids were removed.ii) The fusion protein glycogen synthase from P. abysii was cleaved from the receptor at amino acids Lys294 6.25 and Gln254 5.69 .iii) Loop breaks and fusion protein cleavage sites were capped with methylamine and acetyl at C-and N-termini, respectively.iv) Hydrogen atoms were added, and the protonation states of titratable amino acids were assigned using MOE's Protein Preparation subroutines (pHvalue 7.4, temperature 300 K).Protonation states of histidines and rotamers of head groups of glutamines and asparagines were visually inspected and adjusted according to their protein environment.

Database generation
The DrugsNow subset of the ZINC database (21) was downloaded as SMILES strings (7.377 million entries).The subset was split into batches of 10000 molecules (738 batches).Each batch was submitted to OpenEye's OMEGA program for conformer generation.(22) For each molecular entity, a maximum number of 400 conformers was generated.The energy window parameter was set to 10.0 and the rms (root mean square deviation of conformer coordinates) parameter to 0.5.The "strict" flag was set to "true" ensuring that molecules without fully specified stereochemistry were discarded.A total of 1.726 billion conformers were generated.For docking with DOCK, the ZINC LeadLike subset was downloaded in flexibase format (3.987 million entries) and used as-is (Figure 2A).

Docking
Docking was performed using OpenEye's HYBRID program.(23,24) The ensemble of conformations of each molecule was overlaid with the co-crystallized ligand (suvorexant) in order to determine the best suited conformer for the following exhaustive docking.The method for overlaying conformers is built directly into the HYBRID engine and is based on the same methodology as implemented in the OEChem API and the ROCS application.(25) For the actual docking step -translational and rotational optimization of a compound conformer within the binding site of the protein -HYBRID scores for a given protein-ligand complex were calculated based on the shape and electrostatic complementarity of the ligand and the protein's binding site.Shape and electrostatic features are represented by Gaussian potentials.During optimization, the overlap between ligand and protein features is maximized.After docking, the 100 best-scored poses were extracted from each batch and aggregated into one sorted scoring list (73800, entries, corresponds to best scored percentile of the entire docking run).
The ZINC LeadLike subset was docked with DOCK.( 26) Briefly, molecules were placed using guiding points inside the pocket that had been derived from suvorexant bound to the OX2R in the crystal structure (PDB code 4S0V) (5).
All docking scores can be found in the supplementary information (Table S 4).Molecules were purchased from various vendors, as listed in Table S2.Purity of all ligands is ≥ 95 %, as determined by LC/MS and different methods employed by the vendors.

Post-processing
The scoring list of ~74000 poses was treated in three different ways: i) left unchanged (sorted by HYBRID score), ii) rescored by DSX (27) and iii) submitted to a rigid body optimization procedure using OpenEye's SZYBKI program.(28) During this optimization process, the atoms of the receptor were kept fixed at their crystallographic positions, while the rotational, translational and torsional degrees of freedom of each pose were optimized using a Poisson-Boltzmann solvation model.The resulting poses were sorted by the SZYBKI ligand-protein interaction energy.
From each of the scoring lists (HYBRID, DSX and SZYBKI), a slice of 2000 poses was extracted and subjected to visual inspection in order to remove those that form improbable interactions that are not sufficiently penalized by present-day scoring functions.In addition, the 500 best-scored poses from the docking run with DOCK were added to the visual inspection.Selected compounds were acquired from their respective vendors and analyzed pharmacologically.Parent generation of compounds are denoted with a capital P (Figure 2B).

Similarity screen
Several experimentally validated compounds from the parent generation (both with and without affinity against the receptor) were used as queries for the retrieval of close analogs by a fingerprint-based similarity screen: For each of the nine hits (P7, P9, P21, P22, P27, P31, P32, P33 and P35), the 30 nearest neighbors were retrieved from the three 2000 entries-long slices (HYBRID, DSX and SZYBKI [Figure 2B]).The retrieved poses were all subjected to optimization with SZYBKI.The resulting geometries were visually inspected and the selected compounds were acquired from their respective vendors and analyzed pharmacologically (daughter generation of compounds denoted with a capital F) (Figure 2C).

Experimental validation
Materials 3

Cell-culture
CHO cells stably expressing either the human OX1 or OX2 receptor (originally a gift from Heptares, UK) were secondarily transfected with an SRE-luciferase reporter gene and stable clones selected by dilution cloning to create CHO-OX1 and CHO-OX2 stable cell lines.Cells were grown in Dulbecco's modified Eagle's medium nutrient mix F12 (DMEM/F12) containing 10 % fetal calf serum (FCS) and 2 mM L-glutamine in a 37°C humidified 5 % CO2: 95 % air atmosphere.

Whole cell 3 H-radioligand binding
The affinity of compounds for the OX1 and OX2 receptors were determined from whole cell binding studies.Experiments were conducted on confluent cells in white-sided 96-well view plates as previously described (2 hour 37°C incubation with radioligand and competing ligand in 200 µl serum-free media, Baker 2005).A 7-point concentration response curve (each point in triplicate) was examined for each ligand in each experiment.Suvorexant (10 μM) was used to define non-specific binding.The affinity of the radioligands has previously been determined as 4.65 nM for 3 H-SB674042 in the CHO-OX1 cells and 7.86 nM for 3 H-EMPA in the CHO-OX2 cells.(14) Cells were inspected under a light microscope to ensure they were still present after the 2hour incubation, both before and after the wash.In a few cases, high concentrations (100µM) of competing ligand caused the cells to round up and be washed off the plates.These concentrations were excluded from the analysis.
The IC50 value of competing ligands was determined from a sigmoidal response curve (plotted in Prism 7) where the IC50 is the concentration required to inhibit 50 % of the specific binding of the 3 H-radioligand, [A] is the concentration of the competing ligand.

% 𝑠𝑝𝑒𝑐𝑖𝑓𝑖𝑐 𝑏𝑖𝑛𝑑𝑖𝑛𝑔 = 100 − 100 • [𝐴] 𝐼𝐶 50 + [𝐴]
The Ki values for the competing ligands was then determined from the IC50 values using Cheng-Prusoff equation where [L] is the concentration of radioligand in that experiment and LKD is the KD value of the radioligand (as determined from saturation binding ( 14)):

Intracellular calcium release
Calcium measurements were made using a Flexstation 3 at 37°C.Cells, seeded into black 96-well view plates, were loaded with Fluo-4AM / pluronic-F127 in serum-free media containing probenecid (25 mM) for 45 minutes at 37°C.Cells were washed with 200 μl HEPES buffered saline (HBS, containing 2 μM CaCl2) per well, before a further 80 µl HBS was then added to each well.Ligands were diluted in HBS to 5 times final concentration in round bottomed 96-well compound plates.During the experiment, the Flexstation robotics added 20 µl of ligand from the compound plate into the existing 80 µl HBS in the cell plate (1:5 dilution in well).Ionomycin (10 µM) and orexin A (1 μM) were used as positive controls each plate in each experiment.Calcium mobilization was followed for 120 seconds and data were plotted as the maximum value obtained for calcium mobilization over the basal value obtained for that well before the addition of ligand.
A sigmoidal concentration response curve was fitted to the data (using Prism 7) where Emax is the maximum response, [A] is the agonist concentration and EC50 is the concentration of agonist that produces 50 % of the maximal response:

Associated content
Tabular data as PDF files containing compound structures, SMILES strings, and supplier information; Tanimoto fingerprint similarities, ROCS TanimotoCombo scores, docking scores; and retrosynthetical analysis.Predicted complex structures for all molecules are provided in PDB format.

Table of contents graphic
Table 1: pKi values, obtained for five known orexin ligands and the initial 43 compounds identified from the primary docking screen, from whole cell binding studies as determined from using 3 H-SB674042 (CHO-OX1 cells) and 3 H-EMPA (CHO-OX2 cells).Values are mean ± sem from n separate experiments.app = apparent Ki value.Here, the maximal achievable concentration of competing ligand was not able to fully inhibit the radioligand specific binding.In cases where greater than 50% specific binding was inhibited, an apparent Ki value is given, assuming that if a greater concentration of ligand were possible, full inhibition of specific binding would have occurred.IC50 > 100 µM.Here, inhibition of specific binding by the competing ligand was less than 50% so an IC50 value, and therefore Ki value, could not be calculated.The IC50 must therefore have been greater than the maximum concentration of competing.ep = early plateau.Some ligands did not fully inhibit radioligand binding but appeared to reach an early plateau.Here increasing concentrations of competing ligand reached a maximum inhibition that was less than that achieved by suvorexant (similar to that seen in Proudman and Baker 2018 ( 14)).In these cases, an apparent Ki value is given for the partial inhibition of specific binding.The percentage of specific binding inhibited was 79.2 ± 3.3% for P27 and 63.7 ± 2.1% for P35 in the CHO-OX2 cells.

Figure 2 :
Figure 2: Schematic of the primary docking screen workflow.A: The ZINC DrugsNow subset of ~7.4 M compounds was subjected to conformer generation used for docking.In addition, ~4 M ZINC LeadLike compounds were docked as-is.B: Threefold post-processing of the top 1 % slice (6500 poses -2000 each from HYBRID, DSX, SZYBKI and 500 from DOCK) concluded by visual inspection, selection of compounds and their experimental validation.C: Schematic of the secondary similarity screen workflow.Eight experimentally validated hits were used as queries for a fingerprint-based similarity screen of three 2000 slices from the primary screen.The three sets of 30 nearest neighbors each were optimized with SZYBKI and subjected to visual inspection.Selected virtual hits were validated experimentally.

Figure 3 :
Figure 3: Inhibition of A 3 H-SB67404 binding to CHO-OX1 cells and B 3 H-EMPA binding to CHO-OX2 cells in response to P33, F33.3 and suvorexant.Bars represent total and non-specific binding (as determined by 10 μM suvorexant) and data point are mean ± sem of triplicate determinations.These experiments are representative of 4 separate experiments.The concentration of radioligand in these experiments was A: 1.14 and B: 0.60 nM.

Figure 4 :
Figure 4: The ten compounds with the highest measured affinity in the 3 H-EMPA binding assay to CHO-OX2 cells identified in this study.P33 and P35 compound families are emphasized with a box.

Figure 5 :
Figure 5: Intracellular calcium release in A CHO-OX1 cells and B CHO-OX2 cells in response to orexin A, P33 and F33.3.Bars represent basal intracellular calcium release and that in response to 10µ M ionomycin alone.Data points are mean ± sem of triplicate determinations and these single experiments are representative of 4 separate experiments in each case.

Figure 6 :
Figure 6: Depictions of F33.3 (left) and P33 (right) showing poses of the most potent and the most OX2R selective compound, respectively.Compounds are shown in green sticks.Asn324 6.55 and His350 7.39 in orange sticks (bold and thin, respectively).Helix 7 was removed for clarity.Color code: C green, N blue, O red, S yellow, Cl dark green, F cyan.PDB code: 4S0V.

Figure 7 :
Figure 7: Depictions of poses of P7 (left) and F7.7 (right), respectively.Compounds are shown in green sticks.Asn324 6.55 and His350 7.39 in orange sticks (bold and thin, respectively).Glu212 45.52 and Arg328 6.59 in cyan sticks (bold and thin, respectively).Helix 7 was removed for clarity.Color code: C green, N blue, O red, S yellow, Cl dark green, F cyan.The protein is PDB code: 4S0V.