- Email: [email protected]
Identification of a new small molecule chemotype of Melanin Concentrating Hormone Receptor-1 antagonists using pharmacophore-based virtual screening
Journal Pre-proofs Identification of a New Small Molecule Chemotype of Melanin Concentrating Hormone Receptor-1 Antagonists Using Pharmacophore-Based Virtual Screening Mohamed A. Helal, Amar G. Chittiboyina, Mitchell A. Avery PII: DOI: Reference:
S0960-894X(19)30704-8 https://doi.org/10.1016/j.bmcl.2019.126741 BMCL 126741
To appear in:
Bioorganic & Medicinal Chemistry Letters
Received Date: Revised Date: Accepted Date:
13 June 2019 1 October 2019 3 October 2019
Please cite this article as: Helal, M.A., Chittiboyina, A.G., Avery, M.A., Identification of a New Small Molecule Chemotype of Melanin Concentrating Hormone Receptor-1 Antagonists Using Pharmacophore-Based Virtual Screening, Bioorganic & Medicinal Chemistry Letters (2019), doi: https://doi.org/10.1016/j.bmcl.2019.126741
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
© 2019 Published by Elsevier Ltd.
Identification of a New Small Molecule Chemotype of Melanin Concentrating Hormone Receptor-1 Antagonists Using Pharmacophore-Based Virtual Screening
Mohamed A. Helala,b,*, Amar G. Chittiboyinac, and Mitchell A. Averyd
a
University of Science and technology, Biomedical Sciences Program, Zewail City of Science and Technology,
October Gardens, 6th of October, Giza 12578, Egypt b
Medicinal Chemistry Department, Faculty of Pharmacy, Suez Canal University, Ismailia 41522, Egypt
c
National Center for Natural Products Research, School of Pharmacy, University of Mississippi, University, MS
38677, United States d
Department of BioMolecular Sciences, School of Pharmacy, University of Mississippi, University, MS 38677,
United States
*Corresponding Author: Mohamed A. Helal Tel: +201201122213 Email: [email protected]
1
Abstract MCH receptor is a G protein-coupled receptor with two subtypes R1 and R2. Many studies have demonstrated the role of MCH-R1 in feeding and energy homeostasis. It has been proven that oral administration of small molecule MCH-R1 antagonists significantly reduces food intake and causes a dose-dependent weight loss. In this study, two ligand-based pharmacophores were developed and validated based on recently published MCH-R1 antagonists with diverse structures. Successful pharmacophores had one hydrogen bond acceptor, one positive ionizable, one ring aromatic and two or three hydrophobic groups. These 3D-QSAR models were used for virtual screening of the ZINC chemical database resulting in the identification of nine compounds with more than 50% displacement of radiolabeled MCH at a 20 μM concentration. Moreover, four of these compounds showed antagonistic activities in Aequorin functional assay, including MH-3 which is the first MCH-R1 antagonist based on a diazaspiro[4.5]decane scaffold. The most active compounds were also docked into our previously published MCH-R1 homology model to gain insights into their binding determinants. These compounds could represent a viable starting scaffold for the design of potent MCH-R1 antagonists with improved pharmacokinetic properties as an effective treatment for obesity. Keywords: Melanin concentrating hormone; QSAR; Pharmacophore; Antagonist; spirodecane
2
Obesity is defined as a body mass index (BMI) of 30 or more. The World Health Organization (WHO) formally recognized obesity as a global epidemic in 1997.1 It is a major public health problem affecting both developed and developing countries with the most dramatic impact in urban areas. During early 2000s, Melanin Concentrating Hormone Receptor-1 (MCH-R1) has emerged as a promising target for treatment of obesity. MCH-R1 is involved in the control of energy homeostasis, feeding behavior and body weight.2 The involvement of MCH-R1 in the control of energy expenditure and body weight has motivated scientists from both academia and industry towards the development of MCH-R1 antagonists for the treatment of obesity.3 MCH-R1 is known to bind compounds with diverse chemical structures. However, most of the reported compounds possess a central amide residue with a hydrophobic aromatic group at one end of the molecule and basic nitrogen at the other end. The aromatic group is believed to form πstacking interactions with some aromatic residues within the center of the receptor transmembrane domain, while the basic nitrogen forms an essential salt bridge with the negatively charged Asp192. Due to the lack of X-ray crystal structure data for MCH-R1, the use of ligand-based drug design is a valuable tool for antagonist development.4 Chemical function-based pharmacophores, once validated, are useful in the prediction of the activity of other compounds and also in virtual screening of combinatorial libraries for the discovery of new leads.5 However, there has been only two attempts in the literature, at the time of this study, to develop a pharmacophore model for MCH-R1 antagonists.6, 7 The first one was reported by Ulven et al. from 7TM Pharma. They have used a series of closely related benzamide derivatives synthesized in their laboratory as the training set. Similarly, the second study was based on a single set of compounds and the developed pharmacophore was lacking the “positive ionizable” feature which is known to be essential for MCH-R1 antagonists interaction with the critical aspartic acid residue.
3
In the present study, we have built two ligand-based pharmacophore models utilizing reported MCH-R1 antagonists of highly diverse structures and from independent research laboratories using CATALYST module within Discovery Studio suite (now Biovia).8, 9 Afterwards, these models were validated for their predictive ability using a different set of MCH-R1 antagonists (test set). The developed pharmacophore models were used for virtual screening of a commercial compound database and the hits obtained were subjected to biological evaluation in binding as well as functional assays. We decided to start with developing a HypoGen model. This technique attempts to derive the simplest hypotheses that best correlate the activities of the training set compounds with their chemical structures. The number of compounds in the training set should exceed 16 with a wide range of activity (4-5 orders of magnitude). The compounds should possess structural diversity and contain no redundant information. The resulting hypotheses represent a collection of predefined features (e.g. hydrogen bond acceptor, ring aromatic, positive ionizable) in space. Selection of the training set is the most critical step in pharmacophore generation. The number of compounds in the training set should exceed 16 with a wide range of activity (4-5 orders of magnitude). The compounds should possess structural diversity and contain no redundant information.8 Considering these factors, 32 MCH-R1 antagonists reported by three independent laboratories with activity data (Ki) spanning over 6 orders of magnitude (0.41-14000 nM) were selected to generate the HypoGen hypothesis (Fig. S1a, Supplementary Data).10-18 Depending on the chemical structures of the training set compounds and the earlier MCH pharmacophores, five features, hydrogen bond acceptor (HBA), hydrogen bond donor (HBD), positive ionizable (PI), ring aromatic (RA) and hydrophobic group (HY), were selected from the CATALYST feature dictionary.
4
The initial HypoGen run returned 10 hypotheses. The hydrogen bond donor feature was never used in any hypotheses. In addition, the configuration cost was 16.3 which is a relatively high value. The configuration cost quantifies the complexity of the hypothesis space. It is fixed for all the hypotheses returned from each HypoGen run. If this value is greater than 17, the generated hypotheses may have a chance correlation. This motivated us to perform another HypoGen run with constraints on the number of features as follows, hydrophobic, min 2, max 2; ring aromatic, min 1, max 1; positive ionizable, min 1, max 1; hydrogen bond acceptor, min 1, max 1. The second HypoGen run returned 10 hypotheses; all of them are composed of the five features mentioned above. Figure 1 shows the top scoring hypothesis Hypo1 as well as the correlation coefficient of actual versus estimated activities of the training set. This hypothesis is composed of two hydrophobic features, one ring aromatic, one hydrogen bond acceptor, and one positive ionizable feature. Examination of this hypothesis revealed that the total cost and the configuration cost values of the initial hypotheses were improved. Table S1 (Supplementary data) shows the matrix Distances (in Å) of the chemical features of the best generated HypoGen pharmacophore model. The total cost and Correlation coefficient of the 10 pharmacophore models are shown in Table S2 (Supplementary Data).
5
Figure 1: a) Top scoring HypoGen pharmacophore, Hypo1. Features are color-coded with cyan, yellow, red and green contours representing hydrophobic features, ring aromatic feature, positive ionizable feature and hydrogen bond acceptor feature, respectively. b) The regression of actual versus estimated activities by Hypo1 on the training set. Correlation coefficient = 0.953.
The hypothesis cost value is the number of bits required for complete description of the hypothesis. It is assumed that simpler hypotheses with fewer numbers of bits are preferred. The HypoGen module performs two additional cost calculations, the fixed and the null costs. The fixed cost represents the lowest possible cost for the simplest theoretical model that fits the data perfectly. The null cost is the maximum cost of a pharmacophore with no features that estimates activity to be the average of the training set compounds activity data. The null cost is usually higher than the fixed cost and both of them are reported in the HypoGen log file. For a hypothesis to be statistically significant, a difference of 60 bits should be achieved between the total cost of the hypothesis and the null cost. Hypothesis with cost that differs from the null hypothesis cost by 4060 bits have 75-90% chance of representing true data correlation. The total cost of a hypothesis is calculated as the summation of three components, weight cost, error cost, and configuration cost. The weight of a hypothesis feature describes the activity magnitude represented by this feature. The weight cost increases as the feature weights deviate from an ideal value of 2. The error cost
6
increases with the increase of the root mean square deviation (RMSD) between actual and estimated activities of the training set compounds.8 In the second HypoGen run, the difference between the total cost of the best hypothesis (Hypo1) and the null cost was found to be 65.54, representing more than 90% probability of correlating the data. In addition, a low RMSD value of 0.717 and high correlation coefficient of 0.953 with acceptable configuration cost of 13.649 suggest successful hypothesis generation. Table S3 (Supplementary data) represents the actual and estimated activities of the 32 training set compounds along with the error values (expressed as the ratio between estimated and actual activities), based on the best hypothesis Hypo1. Figure 1c (Supplementary data) shows the mapping of the most potent training set compound (1) onto the best hypothesis, Hypo1. The cyano group maps onto the hydrogen bond acceptor feature, the piperidine nitrogen acts as a positive ionizable group, the western phenyl ring serves as a hydrophobic group and the two phenyl rings adjacent to the piperidine ring map onto the other hydrophobic and the ring aromatic features. In addition to the cost analysis, the best generated HypoGen pharmacophore model (Hypo1) was validated using an external test set as well as the Catscramble technique (Tables S4 and S5, Supplementary data). The test set compounds have activity data (Ki) spanning from 0.17 nM to 17380 nM over a range of six order of magnitude. The best pharmacophore model, Hypo1, was used to estimate the activity of the 30 test set compounds. The test set has a correlation coefficient of 0.833 which indicates a good correlation between actual and estimated activity. Twenty two of the 30 test set compounds were predicted with error value less than 10, representing less than one order difference in activity estimation. The remaining compounds were predicted with error factor not more than 18. Moreover, the hypothesis could discriminate satisfactorily between stereoisomers.
7
Pharmacophore models can be derived from a training set including only the active compounds. This approach is represented by the HipHop method in the Catalyst software. This technique is especially useful when the training set is small and/or the biological data is insufficient.19 In order to generate the HipHop pharmacophore model, ten potent MCH-R1 antagonists, reported by nine different laboratories were selected from the literature (Fig. 2).3, 6, 12, 13, 20-24
The selected MCH-R1 antagonists possess high binding affinity with Ki values ranging
from 0.13 to 0.59 nM. Based on the features characteristics of the training set molecules and the previously generated HypoGen models, the following chemical functions were used for the HipHop model generation: HBA, HBD, PI, RA, and HY. Since the Ki values of the molecules in the training sets are within the sub-nanomolar range, all compounds were considered to contain equally important features for the receptor binding.
Fig. 2: Chemical structures of the 10 training set compounds used to generate the HipHop model.
A total of 10 pharmacophore models were generated by the HipHop hypothesis generation process. Nine of them had five features. The highest ranked hypothesis had six features; one hydrogen bond acceptor, one positive ionizable, one ring aromatic and three hydrophobic groups. This hypothesis was selected for virtual screening because it has the highest number of features
8
that makes it the most selective pharmacophore. By checking the 3D alignment of the ten hypotheses, it was found that they share three features (RA, PI, and HBA). These features show very good overlap in the 3D space, indicating their importance for the binding affinity of the MCHR1 antagonists as well as uniformity of the generated hypotheses. Figure 3a shows the selected MCH-R1 HipHop model. Mapping of the training set compound 39 onto this pharmacophore model is represented in Fig. 3b. The distances among the centers of the features of the best HipHop model are shown in Table S6 (Supplementary Data). This model had been validated using a test of nine compounds (Supplementary Data).
Figure 3: a) HipHop pharmacophore model of the highest rank. PI features are represented by red spheres; RA features are represented by pairs of mashed brown spheres; HY features are represented by cyan spheres; HBA are represented by a pair of green spheres. b) Mapping of the training set compound 39, onto the best HipHop model. Features are color-coded as before.
Having these two validated pharmacophore models in hand, we started our virtual screening campaign. Literature search revealed that MCH-R1 requires lipophilic molecules with high molecular weight. Therefore, we decided to conduct our screening using the “Big” subset in the ZINC online compound database. We have filtered all the compounds based on the following criteria: 3 < logP < 6, number of hydrogen bond acceptors < 10, number of hydrogen bond donors
9
< 5, and number of rotatable bonds < 10. The filtered output was then submitted for virtual screening using the Catsearch background utility. The virtual screening flowchart is summarized in Figure 4. Both HypoGen and HipHop pharmacophores were used individually to screen the Unity prefiltered set (1.45 million compounds). A total of 1811 and 216 hits were retrieved using the HypoGen and HipHop models, respectively. Using a ‘Best fit’ cutoff value of three, hits were filtered to 485 and 159 in case of HypoGen and HipHop, respectively. The combined data set (644 compounds) was visually inspected with respect to perceived chemical stability and toxicity. After checking commercial availability, 24 compounds were purchased and submitted for biological evaluation. Percent displacement of radiolabeled MCH was calculated using membrane preparations from Chinese Hamster Ovary (CHO) cells overexpressing MCH-R1. Membranes obtained from Chinese Hamster Ovary (CHO) cells overexpressing MCHR-1 were used. Radioligand, [125I]MCH, was obtained from Perkin Elmer (NEX 375, 2200 μCi). All components were diluted in assay buffer (25 mM HEPES pH 7.4; 1 mM CaCl2, 5 mM MgCl2, 0.2% protease free BSA), 50 μl of test compound at 20 μM concentration, 25 μl of radioligand, [125I]MCH, diluted in assay buffer, 25 μl of a premix of membrane/SPA beads (PVT-WGA from Amersham Bioscience) were successively added in the wells of an Optiplate (Perkin Elmer). Plates are incubated at room temperature for 2 hours before counting for 1 minute/well in a TopCount (Packard). The data were analyzed by non-linear regression using the GraphPad Prism software. All assays were conducted in duplicates. Nine compounds showed more than 50% displacement of radiolabeled MCH at a 20 μM concentration. Based on the activity and the novelty of the tested compounds, the exact IC50 of five compounds were determined in an Aequorin functional assay. The data of the selected compounds are given in Table 1 and the IC50 curves are shown in Figure 5. As discussed earlier, the most active compounds, MH-1 – MH-3 and MH-5, possess a central
10
amide function with an aromatic ring system at one end and a basic nitrogen carrying another aromatic/hydrophobic group on the other end. The physicochemical descriptors and the pharmacokinetic parameters of the two most active compounds, MH-3 and MH-5, were computed using the SwissADME server in order to evaluate their druglike nature.25 These compounds showed favorable physicochemical properties with molecular weights around 400 and LogP less than 4.2. Both compounds were predicted to show a high GI absorption and BBB penetration with all parameters lying within the physicochemical space of the SwissADME hexagonal model for oral bioavailability (Fig. S4). This would facilitate their SAR exploration while keeping the compounds favorable physicochemical properties. Concerning the compounds novelty, the piperidine nucleus were previously exploited for the design of MCH-R1 antagonists. However, the 2,7-diazaspiro[4.5]decane scaffold of compound MH-3 has never been used as a template for MCH-R1 ligands. Interestingly, the most active compounds obtained from virtual screening, including the spiro[4.5]decane MH-3, showed a three-carbon distance between the essential basic nitrogen and the amide function as well as a benzyl substituent on the basic nitrogen. This would allow us to set a uniform systematic lead optimization campaign.
11
Figure 4: The protocol utilized for the virtual screening study.
Figure 5: Aequorin functional assay of the top five compounds, using CHO cell lines overexpressing MCH-R1.
Table 1: The ID number, vendor, structure, and activity data of the obtained leads. ID
Vendor
MH-1
Chembridge
Structure
Mol. Wt.
Log P
% Displacement* (IC50)‡
372.89
4.02
79.83 (23 μM)
12
MH-2
Enamine
427.33
4.44
71.36 (23 μM)
MH-3
Chembridge
408.53
4.27
73.85 (18 μM)
MH-4
Chembridge
312.38
3.05
65.56 (43 μM)
MH-5
Chembridge
384.49
4.49
88.99 (11 μM)
MH-6
Chembridge
425.59
6.22
58.18
MH-7
Enamine
385.50
4.45
90.58
MH-8
Enamine
428.94
3.12
52.88
MH-9
Chembridge
425.35
5.75
58.28
* Displacement of [125I]MCH at inhibitor concentration of 20 μM, all values were determined in duplicate. The reference compound used is ATC-0175 which showed an IC50 of 8.28 nM. ‡ IC values were calculated in Aequorine functional assay, all values were determined in duplicate. 50
In order to understand the binding determinants of the discovered lead compounds and set the stage for the future optimization efforts, we have docked the most active compounds, MH-3 and MH-5, into a homology model of MCH-R1. This model was built based on the highest resolution crystal structure of bovine rhodopsin (PDB ID: 1U19). The crude model was refined using OPLS2005 force field in several rounds of consecutive minimization and gradually reducing constraints. Then, the minimized model was inserted into a pre-equilibrated DPPC/TIP3P membrane aqueous system and subjected to a ligand-steered refinement protocol using a 20 ns molecular dynamics simulation in an NPT ensemble.26 It was clear that both molecules docked into the receptor
13
orthosteric binding site formed by TMIII, TMV, TMVI, and TMVII, showing the essential salt bridge with Asp192 using their basic nitrogens (Fig. 6a-d). The aryl groups directly attached to the amide linker were found deeply inserted into the hydrophobic pocket formed by Tyr342, Phe286, Phe290, and Leu203, while the benzyl moiety on the basic nitrogen was directed upwards making contacts with the residues of the upper part of the binding site. Despite the presence of the critical ionic bond with Asp192, interactions with the hydrophobic residues are still not ideal, which is expected at the micromolar affinity level. Based on this observation, we envisioned a future optimization campaign of two main stages. First, the top benzyl moiety, which is solvent exposed, should be substituted with polar groups allowing more favorable contacts with water molecules and\or nearby polar residues such as Gln345. In the second stage, a short flexible linker could be inserted between the central amide function and the lower aryl group. This might allow the methoxy and fluoro groups, or any other polar substituents, to form H bonds with Thr200 at the bottom of the binding site. Moreover, the flexibility of the linker would allow this aryl ring to form π-stacking interactions with Phe286 or Phe290. In addition, it was assumed that MCH-R1 antagonists of all chemotypes can form H bonds with the side chain of Gln196. This came in accordance with our pharmacophore models which include a central HBA feature. Introducing a short flexible linker to compounds MH-3 and MH5 could help the central amide group to form this H bonding interaction with Gln196. Noteworthy, this is the first report of the utility of spiro[4.5]decane as a template for MCH-R1 antagonists (compound MH-3). Building on this scaffold, we hope to reduce the aromatic nature of our MCHR1 lead compounds to achieve better pharmacokinetics. This will be the objective of our future endeavors. Optimization of these compounds could eventually lead to the discovery of an efficient and safe treatment of obesity and a relief for its associated complications.
14
Figure 6: Proposed binding modes of MH-3 (a and b) and MH-5 (c and d) in our previously published homology model of MCH-R1. Ligands are shown as brown sticks, while residues within 5 Å are shown as cyan sticks. Polar interactions are displayed as red dashed lines. In the 2D interaction maps, residues involved in hydrophobic and polar interaction are represented by green and pink circles, respectively. Polar interactions with Asp192 are represented by a pink dashed line.
References 1. Katzmarzyk, P. T. ;Chaput, J. P. ;Fogelholm, M. ;Hu, G. ;Maher, C. ;Maia, J. ;Olds, T. ;Sarmiento, O. L. ;Standage, M. ;Tremblay, M. S. ;Tudor-Locke, C., International Study of Childhood Obesity, Lifestyle and the Environment (ISCOLE): Contributions to Understanding the Global Obesity Epidemic. Nutrients 2019, 11(4), 848. 2. Sabetghadam, A.; Grabowiecka-Nowak, A.; Kania, A.; Gugula, A.; Blasiak, E.; Blasiak, T.; Ma, S.; Gundlach, A. L.; Blasiak, A., Melanin-concentrating hormone and orexin systems in rat nucleus incertus: Dual innervation, bidirectional effects on neuron activity, and differential influences on arousal and feeding. Neuropharmacology 2018, 139, 238-256. 3. Kawata, Y. ;Okuda, S. ;Hotta, N. ;Igawa, H. ;Takahashi, M. ;Ikoma, M. ;Kasai, S. ;Ando, A. ;Satomi, Y. ;Nishida, M. ;Nakayama, M. ;Yamamoto, S. ;Nagisa, Y. ;Takekawa, S., A novel
15
and selective melanin-concentrating hormone receptor 1 antagonist ameliorates obesity and hepatic steatosis in diet-induced obese rodent models. Eur J Pharmacol 2017, 796, 45-53. 4. Lu, X.; Yang, H.; Chen, Y.; Li, Q.; He, S. Y.; Jiang, X.; Feng, F.; Qu, W.; Sun, H., The Development of Pharmacophore Modeling: Generation and Recent Applications in Drug Discovery. Curr Pharm Des 2018, 24(29), 3424-3439. 5. Dash, R. C. ;Ozen, Z. ;McCarthy, K. R. ;Chatterjee, N. ;Harris, C. A. ;Rizzo, A. A. ;Walker, G. C. ;Korzhnev, D. M. ;Hadden, M. K., Virtual Pharmacophore Screening Identifies SmallMolecule Inhibitors of the Rev1-CT/RIR Protein-Protein Interaction. ChemMedChem 2019, 14 (17), 1610-17. 6. Ulven, T.; Frimurer, T. M.; Receveur, J. M.; Little, P. B.; Rist, O.; Norregaard, P. K.; Hogberg, T., 6-Acylamino-2-aminoquinolines as potent melanin-concentrating hormone 1 receptor antagonists. Identification, structure-activity relationship, and investigation of binding mode. J Med Chem 2005, 48 (18), 5684-97. 7. Abu-Hammad, A.; Zalloum, W. A.; Zalloum, H.; Abu-Sheikha, G.; Taha, M. O., Homology modeling of MCH1 receptor and validation by docking/scoring and protein-aligned CoMFA. Eur J Med Chem 2009, 44 (6), 2583-96. 8. CATALYST 4.8 User Guide, A. I., San Diego, CA, USA, 2003. 9. Kurogi, Y.; Guner, O. F., Pharmacophore modeling and three-dimensional database searching for drug design using catalyst. Curr Med Chem 2001, 8 (9), 1035-55. 10. Guo, T.; Hunter, R. C.; Gu, H.; Rokosz, L. L.; Stauffer, T. M.; Hobbs, D. W., Discovery and SAR of 4-amino-2-biarylbutylurea MCH 1 receptor antagonists through solid-phase parallel synthesis. Bioorg Med Chem Lett 2005, 15 (16), 3691-5. 11. Su, J.; McKittrick, B. A.; Tang, H.; Czarniecki, M.; Greenlee, W. J.; Hawes, B. E.; O'Neill, K., Discovery of melanin-concentrating hormone receptor R1 antagonists using highthroughput synthesis. Bioorg Med Chem 2005, 13 (5), 1829-36. 12. Guo, T.; Shao, Y.; Qian, G.; Rokosz, L. L.; Stauffer, T. M.; Hunter, R. C.; Babu, S. D.; Gu, H.; Hobbs, D. W., Discovery and SAR of biaryl piperidine MCH1 receptor antagonists through solid-phase encoded combinatorial synthesis. Bioorg Med Chem Lett 2005, 15 (16), 3696-700. 13. Wu, W. L.; Burnett, D. A.; Caplen, M. A.; Domalski, M. S.; Bennett, C.; Greenlee, W. J.; Hawes, B. E.; O'Neill, K.; Weig, B.; Weston, D.; Spar, B.; Kowalski, T., Design and synthesis of orally efficacious benzimidazoles as melanin-concentrating hormone receptor 1 antagonists. Bioorg Med Chem Lett 2006, 16 (14), 3674-8. 14. Sasikumar, T. K.; Qiang, L.; Burnett, D. A.; Greenlee, W. J.; Hawes, B. E.; Kowalski, T. J.; O'Neill, K.; Spar, B. D.; Weig, B., Novel aminobenzimidazoles as selective MCH-R1 antagonists for the treatment of metabolic diseases. Bioorg Med Chem Lett 2006, 16 (20), 5427-31. 15. Meyers, K. M.; Mendez-Andino, J.; Colson, A. O.; Hu, X. E.; Wos, J. A.; Mitchell, M. C.; Hodge, K.; Howard, J.; Paris, J. L.; Dowty, M. E.; Obringer, C. M.; Reizes, O., Novel pyrazolopiperazinone- and pyrrolopiperazinone-based MCH-R1 antagonists. Bioorg Med Chem Lett 2007, 17 (3), 657-61. 16. Meyers, K. M.; Kim, N.; Mendez-Andino, J. L.; Hu, X. E.; Mumin, R. N.; Klopfenstein, S. R.; Wos, J. A.; Mitchell, M. C.; Paris, J. L.; Ackley, D. C.; Holbert, J. K.; Mittelstadt, S. W.; Reizes, O., Aminomethyl tetrahydronaphthalene biphenyl carboxamide MCH-R1 antagonists--Increasing selectivity over hERG. Bioorg Med Chem Lett 2007, 17 (3), 814-8.
16
17. Kamata, M.; Yamashita, T.; Imaeda, T.; Tanaka, T.; Masada, S.; Kamaura, M.; Kasai, S.; Hara, R.; Sasaki, S.; Takekawa, S.; Asami, A.; Kaisho, T.; Suzuki, N.; Ashina, S.; Ogino, H.; Nakano, Y.; Nagisa, Y.; Kato, K.; Ishihara, Y., Melanin-concentrating hormone receptor 1 antagonists. Synthesis and structure-activity relationships of novel 3(aminomethyl)quinolines. J Med Chem 2012, 55 (5), 2353-66. 18. Oost, T.; Heckel, A.; Kley, J. T.; Lehmann, T.; Müller, S.; Roth, G. J.; Rudolf, K.; Arndt, K.; Budzinski, R.; Lenter, M.; Lotz, R. R.; Maier, G. M.; Markert, M.; Thomas, L.; Stenkamp, D., Design, synthesis and evaluation of MCH receptor 1 antagonists--Part III: Discovery of pre-clinical development candidate BI 186908. Bioorg Med Chem Lett 2015, 25 (16), 3275-80. 19. Guner, O. F., Pharmacophore perception, development, and use in drug design. 1st ed.; International University Line: San Diego, CA, 2000. 20. Zhang, M.; Tamiya, J.; Nguyen, L.; Rowbottom, M. W.; Dyck, B.; Vickers, T. D.; Grey, J.; Schwarz, D. A.; Heise, C. E.; Haelewyn, J.; Mistry, M. S.; Goodfellow, V. S., Thienopyrimidinone bis-aminopyrrolidine ureas as potent melanin-concentrating hormone receptor-1 (MCH-R1) antagonists. Bioorg Med Chem Lett 2007, 17 (9), 2535-9. 21. Judd, A. S.; Souers, A. J.; Wodka, D.; Zhao, G.; Mulhern, M. M.; Iyengar, R. R.; Gao, J.; Lynch, J. K.; Freeman, J. C.; Falls, H. D.; Brodjian, S.; Dayton, B. D.; Reilly, R. M.; Gintant, G.; Limberis, J. T.; Mikhail, A.; Leitza, S. T.; Houseman, K. A.; Diaz, G.; Bush, E. N.; Shapiro, R.; Knourek-Segel, V.; Hernandez, L. E.; Marsh, K. C.; Sham, H. L.; Collins, C. A.; Kym, P. R., Identification of diamino chromone-2-carboxamides as MCHr1 antagonists with minimal hERG channel activity. Bioorg Med Chem Lett 2007, 17 (8), 236571. 22. Hudson, S.; Kiankarimi, M.; Rowbottom, M. W.; Vickers, T. D.; Wu, D.; Pontillo, J.; Ching, B.; Dwight, W.; Goodfellow, V. S.; Schwarz, D.; Heise, C. E.; Madan, A.; Wen, J.; Ban, W.; Wang, H.; Wade, W. S., Synthesis and structure-activity relationships of retro bisaminopyrrolidine urea (rAPU) derived small-molecule antagonists of the melaninconcentrating hormone receptor-1 (MCH-R1). Part 2. Bioorg Med Chem Lett 2006, 16 (18), 4922-30. 23. Su, J.; McKittrick, B. A.; Tang, H.; Burnett, D. A.; Clader, J. W.; Greenlee, W. J.; Hawes, B. E.; O'Neill, K.; Spar, B.; Weig, B.; Kowalski, T.; Sorota, S.; Li, C.; Liu, T., SAR study of bicyclo[4.1.0]heptanes as melanin-concentrating hormone receptor R1 antagonists: taming hERG. Bioorg Med Chem 2007, 15 (16), 5369-85. 24. Johansson, A., Evolution of physicochemical properties of melanin concentrating hormone receptor 1 (MCHr1) antagonists. Bioorg Med Chem Lett 2016, 26 (19), 4559-4564. 25. Daina, A.; Michielin, O.; Zoete, V., SwissADME: a free web tool to evaluate pharmacokinetics, drug-likeness and medicinal chemistry friendliness of small molecules. Sci Rep 2017, 7, 42717. 26. Helal, M. A.; Chittiboyina, A. G.; Avery, M. A., New insights into the binding mode of melanin concentrating hormone receptor-1 antagonists: homology modeling and explicit membrane molecular dynamics simulation study. J Chem Inf Model 2011, 51 (3), 635-46.
17
Highlights
Melanin Concentrating Hormone Receptor-1 is implicated in the regulation of food intake and energy homeostasis.
Two 3D pharmacophore models for MCHR-1 antagonists have been developed and validated.
The ZINC compound database was screened using the best models obtained.
Nine lead compounds showed more than 50% displacement of radiolabeled MCH at a 20 μM concentration.
The most active compounds were docked into a homology model of MCHR-1 to gain insights into their binding determinants.
18
19
20
21
22
Declaration of interests ☒ The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. ☐The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:
23
Mohamed Helal October 1st, 2019
24
S0960-894X(19)30704-8 https://doi.org/10.1016/j.bmcl.2019.126741 BMCL 126741
To appear in:
Bioorganic & Medicinal Chemistry Letters
Received Date: Revised Date: Accepted Date:
13 June 2019 1 October 2019 3 October 2019
Please cite this article as: Helal, M.A., Chittiboyina, A.G., Avery, M.A., Identification of a New Small Molecule Chemotype of Melanin Concentrating Hormone Receptor-1 Antagonists Using Pharmacophore-Based Virtual Screening, Bioorganic & Medicinal Chemistry Letters (2019), doi: https://doi.org/10.1016/j.bmcl.2019.126741
This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
© 2019 Published by Elsevier Ltd.
Identification of a New Small Molecule Chemotype of Melanin Concentrating Hormone Receptor-1 Antagonists Using Pharmacophore-Based Virtual Screening
Mohamed A. Helala,b,*, Amar G. Chittiboyinac, and Mitchell A. Averyd
a
University of Science and technology, Biomedical Sciences Program, Zewail City of Science and Technology,
October Gardens, 6th of October, Giza 12578, Egypt b
Medicinal Chemistry Department, Faculty of Pharmacy, Suez Canal University, Ismailia 41522, Egypt
c
National Center for Natural Products Research, School of Pharmacy, University of Mississippi, University, MS
38677, United States d
Department of BioMolecular Sciences, School of Pharmacy, University of Mississippi, University, MS 38677,
United States
*Corresponding Author: Mohamed A. Helal Tel: +201201122213 Email: [email protected]
1
Abstract MCH receptor is a G protein-coupled receptor with two subtypes R1 and R2. Many studies have demonstrated the role of MCH-R1 in feeding and energy homeostasis. It has been proven that oral administration of small molecule MCH-R1 antagonists significantly reduces food intake and causes a dose-dependent weight loss. In this study, two ligand-based pharmacophores were developed and validated based on recently published MCH-R1 antagonists with diverse structures. Successful pharmacophores had one hydrogen bond acceptor, one positive ionizable, one ring aromatic and two or three hydrophobic groups. These 3D-QSAR models were used for virtual screening of the ZINC chemical database resulting in the identification of nine compounds with more than 50% displacement of radiolabeled MCH at a 20 μM concentration. Moreover, four of these compounds showed antagonistic activities in Aequorin functional assay, including MH-3 which is the first MCH-R1 antagonist based on a diazaspiro[4.5]decane scaffold. The most active compounds were also docked into our previously published MCH-R1 homology model to gain insights into their binding determinants. These compounds could represent a viable starting scaffold for the design of potent MCH-R1 antagonists with improved pharmacokinetic properties as an effective treatment for obesity. Keywords: Melanin concentrating hormone; QSAR; Pharmacophore; Antagonist; spirodecane
2
Obesity is defined as a body mass index (BMI) of 30 or more. The World Health Organization (WHO) formally recognized obesity as a global epidemic in 1997.1 It is a major public health problem affecting both developed and developing countries with the most dramatic impact in urban areas. During early 2000s, Melanin Concentrating Hormone Receptor-1 (MCH-R1) has emerged as a promising target for treatment of obesity. MCH-R1 is involved in the control of energy homeostasis, feeding behavior and body weight.2 The involvement of MCH-R1 in the control of energy expenditure and body weight has motivated scientists from both academia and industry towards the development of MCH-R1 antagonists for the treatment of obesity.3 MCH-R1 is known to bind compounds with diverse chemical structures. However, most of the reported compounds possess a central amide residue with a hydrophobic aromatic group at one end of the molecule and basic nitrogen at the other end. The aromatic group is believed to form πstacking interactions with some aromatic residues within the center of the receptor transmembrane domain, while the basic nitrogen forms an essential salt bridge with the negatively charged Asp192. Due to the lack of X-ray crystal structure data for MCH-R1, the use of ligand-based drug design is a valuable tool for antagonist development.4 Chemical function-based pharmacophores, once validated, are useful in the prediction of the activity of other compounds and also in virtual screening of combinatorial libraries for the discovery of new leads.5 However, there has been only two attempts in the literature, at the time of this study, to develop a pharmacophore model for MCH-R1 antagonists.6, 7 The first one was reported by Ulven et al. from 7TM Pharma. They have used a series of closely related benzamide derivatives synthesized in their laboratory as the training set. Similarly, the second study was based on a single set of compounds and the developed pharmacophore was lacking the “positive ionizable” feature which is known to be essential for MCH-R1 antagonists interaction with the critical aspartic acid residue.
3
In the present study, we have built two ligand-based pharmacophore models utilizing reported MCH-R1 antagonists of highly diverse structures and from independent research laboratories using CATALYST module within Discovery Studio suite (now Biovia).8, 9 Afterwards, these models were validated for their predictive ability using a different set of MCH-R1 antagonists (test set). The developed pharmacophore models were used for virtual screening of a commercial compound database and the hits obtained were subjected to biological evaluation in binding as well as functional assays. We decided to start with developing a HypoGen model. This technique attempts to derive the simplest hypotheses that best correlate the activities of the training set compounds with their chemical structures. The number of compounds in the training set should exceed 16 with a wide range of activity (4-5 orders of magnitude). The compounds should possess structural diversity and contain no redundant information. The resulting hypotheses represent a collection of predefined features (e.g. hydrogen bond acceptor, ring aromatic, positive ionizable) in space. Selection of the training set is the most critical step in pharmacophore generation. The number of compounds in the training set should exceed 16 with a wide range of activity (4-5 orders of magnitude). The compounds should possess structural diversity and contain no redundant information.8 Considering these factors, 32 MCH-R1 antagonists reported by three independent laboratories with activity data (Ki) spanning over 6 orders of magnitude (0.41-14000 nM) were selected to generate the HypoGen hypothesis (Fig. S1a, Supplementary Data).10-18 Depending on the chemical structures of the training set compounds and the earlier MCH pharmacophores, five features, hydrogen bond acceptor (HBA), hydrogen bond donor (HBD), positive ionizable (PI), ring aromatic (RA) and hydrophobic group (HY), were selected from the CATALYST feature dictionary.
4
The initial HypoGen run returned 10 hypotheses. The hydrogen bond donor feature was never used in any hypotheses. In addition, the configuration cost was 16.3 which is a relatively high value. The configuration cost quantifies the complexity of the hypothesis space. It is fixed for all the hypotheses returned from each HypoGen run. If this value is greater than 17, the generated hypotheses may have a chance correlation. This motivated us to perform another HypoGen run with constraints on the number of features as follows, hydrophobic, min 2, max 2; ring aromatic, min 1, max 1; positive ionizable, min 1, max 1; hydrogen bond acceptor, min 1, max 1. The second HypoGen run returned 10 hypotheses; all of them are composed of the five features mentioned above. Figure 1 shows the top scoring hypothesis Hypo1 as well as the correlation coefficient of actual versus estimated activities of the training set. This hypothesis is composed of two hydrophobic features, one ring aromatic, one hydrogen bond acceptor, and one positive ionizable feature. Examination of this hypothesis revealed that the total cost and the configuration cost values of the initial hypotheses were improved. Table S1 (Supplementary data) shows the matrix Distances (in Å) of the chemical features of the best generated HypoGen pharmacophore model. The total cost and Correlation coefficient of the 10 pharmacophore models are shown in Table S2 (Supplementary Data).
5
Figure 1: a) Top scoring HypoGen pharmacophore, Hypo1. Features are color-coded with cyan, yellow, red and green contours representing hydrophobic features, ring aromatic feature, positive ionizable feature and hydrogen bond acceptor feature, respectively. b) The regression of actual versus estimated activities by Hypo1 on the training set. Correlation coefficient = 0.953.
The hypothesis cost value is the number of bits required for complete description of the hypothesis. It is assumed that simpler hypotheses with fewer numbers of bits are preferred. The HypoGen module performs two additional cost calculations, the fixed and the null costs. The fixed cost represents the lowest possible cost for the simplest theoretical model that fits the data perfectly. The null cost is the maximum cost of a pharmacophore with no features that estimates activity to be the average of the training set compounds activity data. The null cost is usually higher than the fixed cost and both of them are reported in the HypoGen log file. For a hypothesis to be statistically significant, a difference of 60 bits should be achieved between the total cost of the hypothesis and the null cost. Hypothesis with cost that differs from the null hypothesis cost by 4060 bits have 75-90% chance of representing true data correlation. The total cost of a hypothesis is calculated as the summation of three components, weight cost, error cost, and configuration cost. The weight of a hypothesis feature describes the activity magnitude represented by this feature. The weight cost increases as the feature weights deviate from an ideal value of 2. The error cost
6
increases with the increase of the root mean square deviation (RMSD) between actual and estimated activities of the training set compounds.8 In the second HypoGen run, the difference between the total cost of the best hypothesis (Hypo1) and the null cost was found to be 65.54, representing more than 90% probability of correlating the data. In addition, a low RMSD value of 0.717 and high correlation coefficient of 0.953 with acceptable configuration cost of 13.649 suggest successful hypothesis generation. Table S3 (Supplementary data) represents the actual and estimated activities of the 32 training set compounds along with the error values (expressed as the ratio between estimated and actual activities), based on the best hypothesis Hypo1. Figure 1c (Supplementary data) shows the mapping of the most potent training set compound (1) onto the best hypothesis, Hypo1. The cyano group maps onto the hydrogen bond acceptor feature, the piperidine nitrogen acts as a positive ionizable group, the western phenyl ring serves as a hydrophobic group and the two phenyl rings adjacent to the piperidine ring map onto the other hydrophobic and the ring aromatic features. In addition to the cost analysis, the best generated HypoGen pharmacophore model (Hypo1) was validated using an external test set as well as the Catscramble technique (Tables S4 and S5, Supplementary data). The test set compounds have activity data (Ki) spanning from 0.17 nM to 17380 nM over a range of six order of magnitude. The best pharmacophore model, Hypo1, was used to estimate the activity of the 30 test set compounds. The test set has a correlation coefficient of 0.833 which indicates a good correlation between actual and estimated activity. Twenty two of the 30 test set compounds were predicted with error value less than 10, representing less than one order difference in activity estimation. The remaining compounds were predicted with error factor not more than 18. Moreover, the hypothesis could discriminate satisfactorily between stereoisomers.
7
Pharmacophore models can be derived from a training set including only the active compounds. This approach is represented by the HipHop method in the Catalyst software. This technique is especially useful when the training set is small and/or the biological data is insufficient.19 In order to generate the HipHop pharmacophore model, ten potent MCH-R1 antagonists, reported by nine different laboratories were selected from the literature (Fig. 2).3, 6, 12, 13, 20-24
The selected MCH-R1 antagonists possess high binding affinity with Ki values ranging
from 0.13 to 0.59 nM. Based on the features characteristics of the training set molecules and the previously generated HypoGen models, the following chemical functions were used for the HipHop model generation: HBA, HBD, PI, RA, and HY. Since the Ki values of the molecules in the training sets are within the sub-nanomolar range, all compounds were considered to contain equally important features for the receptor binding.
Fig. 2: Chemical structures of the 10 training set compounds used to generate the HipHop model.
A total of 10 pharmacophore models were generated by the HipHop hypothesis generation process. Nine of them had five features. The highest ranked hypothesis had six features; one hydrogen bond acceptor, one positive ionizable, one ring aromatic and three hydrophobic groups. This hypothesis was selected for virtual screening because it has the highest number of features
8
that makes it the most selective pharmacophore. By checking the 3D alignment of the ten hypotheses, it was found that they share three features (RA, PI, and HBA). These features show very good overlap in the 3D space, indicating their importance for the binding affinity of the MCHR1 antagonists as well as uniformity of the generated hypotheses. Figure 3a shows the selected MCH-R1 HipHop model. Mapping of the training set compound 39 onto this pharmacophore model is represented in Fig. 3b. The distances among the centers of the features of the best HipHop model are shown in Table S6 (Supplementary Data). This model had been validated using a test of nine compounds (Supplementary Data).
Figure 3: a) HipHop pharmacophore model of the highest rank. PI features are represented by red spheres; RA features are represented by pairs of mashed brown spheres; HY features are represented by cyan spheres; HBA are represented by a pair of green spheres. b) Mapping of the training set compound 39, onto the best HipHop model. Features are color-coded as before.
Having these two validated pharmacophore models in hand, we started our virtual screening campaign. Literature search revealed that MCH-R1 requires lipophilic molecules with high molecular weight. Therefore, we decided to conduct our screening using the “Big” subset in the ZINC online compound database. We have filtered all the compounds based on the following criteria: 3 < logP < 6, number of hydrogen bond acceptors < 10, number of hydrogen bond donors
9
< 5, and number of rotatable bonds < 10. The filtered output was then submitted for virtual screening using the Catsearch background utility. The virtual screening flowchart is summarized in Figure 4. Both HypoGen and HipHop pharmacophores were used individually to screen the Unity prefiltered set (1.45 million compounds). A total of 1811 and 216 hits were retrieved using the HypoGen and HipHop models, respectively. Using a ‘Best fit’ cutoff value of three, hits were filtered to 485 and 159 in case of HypoGen and HipHop, respectively. The combined data set (644 compounds) was visually inspected with respect to perceived chemical stability and toxicity. After checking commercial availability, 24 compounds were purchased and submitted for biological evaluation. Percent displacement of radiolabeled MCH was calculated using membrane preparations from Chinese Hamster Ovary (CHO) cells overexpressing MCH-R1. Membranes obtained from Chinese Hamster Ovary (CHO) cells overexpressing MCHR-1 were used. Radioligand, [125I]MCH, was obtained from Perkin Elmer (NEX 375, 2200 μCi). All components were diluted in assay buffer (25 mM HEPES pH 7.4; 1 mM CaCl2, 5 mM MgCl2, 0.2% protease free BSA), 50 μl of test compound at 20 μM concentration, 25 μl of radioligand, [125I]MCH, diluted in assay buffer, 25 μl of a premix of membrane/SPA beads (PVT-WGA from Amersham Bioscience) were successively added in the wells of an Optiplate (Perkin Elmer). Plates are incubated at room temperature for 2 hours before counting for 1 minute/well in a TopCount (Packard). The data were analyzed by non-linear regression using the GraphPad Prism software. All assays were conducted in duplicates. Nine compounds showed more than 50% displacement of radiolabeled MCH at a 20 μM concentration. Based on the activity and the novelty of the tested compounds, the exact IC50 of five compounds were determined in an Aequorin functional assay. The data of the selected compounds are given in Table 1 and the IC50 curves are shown in Figure 5. As discussed earlier, the most active compounds, MH-1 – MH-3 and MH-5, possess a central
10
amide function with an aromatic ring system at one end and a basic nitrogen carrying another aromatic/hydrophobic group on the other end. The physicochemical descriptors and the pharmacokinetic parameters of the two most active compounds, MH-3 and MH-5, were computed using the SwissADME server in order to evaluate their druglike nature.25 These compounds showed favorable physicochemical properties with molecular weights around 400 and LogP less than 4.2. Both compounds were predicted to show a high GI absorption and BBB penetration with all parameters lying within the physicochemical space of the SwissADME hexagonal model for oral bioavailability (Fig. S4). This would facilitate their SAR exploration while keeping the compounds favorable physicochemical properties. Concerning the compounds novelty, the piperidine nucleus were previously exploited for the design of MCH-R1 antagonists. However, the 2,7-diazaspiro[4.5]decane scaffold of compound MH-3 has never been used as a template for MCH-R1 ligands. Interestingly, the most active compounds obtained from virtual screening, including the spiro[4.5]decane MH-3, showed a three-carbon distance between the essential basic nitrogen and the amide function as well as a benzyl substituent on the basic nitrogen. This would allow us to set a uniform systematic lead optimization campaign.
11
Figure 4: The protocol utilized for the virtual screening study.
Figure 5: Aequorin functional assay of the top five compounds, using CHO cell lines overexpressing MCH-R1.
Table 1: The ID number, vendor, structure, and activity data of the obtained leads. ID
Vendor
MH-1
Chembridge
Structure
Mol. Wt.
Log P
% Displacement* (IC50)‡
372.89
4.02
79.83 (23 μM)
12
MH-2
Enamine
427.33
4.44
71.36 (23 μM)
MH-3
Chembridge
408.53
4.27
73.85 (18 μM)
MH-4
Chembridge
312.38
3.05
65.56 (43 μM)
MH-5
Chembridge
384.49
4.49
88.99 (11 μM)
MH-6
Chembridge
425.59
6.22
58.18
MH-7
Enamine
385.50
4.45
90.58
MH-8
Enamine
428.94
3.12
52.88
MH-9
Chembridge
425.35
5.75
58.28
* Displacement of [125I]MCH at inhibitor concentration of 20 μM, all values were determined in duplicate. The reference compound used is ATC-0175 which showed an IC50 of 8.28 nM. ‡ IC values were calculated in Aequorine functional assay, all values were determined in duplicate. 50
In order to understand the binding determinants of the discovered lead compounds and set the stage for the future optimization efforts, we have docked the most active compounds, MH-3 and MH-5, into a homology model of MCH-R1. This model was built based on the highest resolution crystal structure of bovine rhodopsin (PDB ID: 1U19). The crude model was refined using OPLS2005 force field in several rounds of consecutive minimization and gradually reducing constraints. Then, the minimized model was inserted into a pre-equilibrated DPPC/TIP3P membrane aqueous system and subjected to a ligand-steered refinement protocol using a 20 ns molecular dynamics simulation in an NPT ensemble.26 It was clear that both molecules docked into the receptor
13
orthosteric binding site formed by TMIII, TMV, TMVI, and TMVII, showing the essential salt bridge with Asp192 using their basic nitrogens (Fig. 6a-d). The aryl groups directly attached to the amide linker were found deeply inserted into the hydrophobic pocket formed by Tyr342, Phe286, Phe290, and Leu203, while the benzyl moiety on the basic nitrogen was directed upwards making contacts with the residues of the upper part of the binding site. Despite the presence of the critical ionic bond with Asp192, interactions with the hydrophobic residues are still not ideal, which is expected at the micromolar affinity level. Based on this observation, we envisioned a future optimization campaign of two main stages. First, the top benzyl moiety, which is solvent exposed, should be substituted with polar groups allowing more favorable contacts with water molecules and\or nearby polar residues such as Gln345. In the second stage, a short flexible linker could be inserted between the central amide function and the lower aryl group. This might allow the methoxy and fluoro groups, or any other polar substituents, to form H bonds with Thr200 at the bottom of the binding site. Moreover, the flexibility of the linker would allow this aryl ring to form π-stacking interactions with Phe286 or Phe290. In addition, it was assumed that MCH-R1 antagonists of all chemotypes can form H bonds with the side chain of Gln196. This came in accordance with our pharmacophore models which include a central HBA feature. Introducing a short flexible linker to compounds MH-3 and MH5 could help the central amide group to form this H bonding interaction with Gln196. Noteworthy, this is the first report of the utility of spiro[4.5]decane as a template for MCH-R1 antagonists (compound MH-3). Building on this scaffold, we hope to reduce the aromatic nature of our MCHR1 lead compounds to achieve better pharmacokinetics. This will be the objective of our future endeavors. Optimization of these compounds could eventually lead to the discovery of an efficient and safe treatment of obesity and a relief for its associated complications.
14
Figure 6: Proposed binding modes of MH-3 (a and b) and MH-5 (c and d) in our previously published homology model of MCH-R1. Ligands are shown as brown sticks, while residues within 5 Å are shown as cyan sticks. Polar interactions are displayed as red dashed lines. In the 2D interaction maps, residues involved in hydrophobic and polar interaction are represented by green and pink circles, respectively. Polar interactions with Asp192 are represented by a pink dashed line.
References 1. Katzmarzyk, P. T. ;Chaput, J. P. ;Fogelholm, M. ;Hu, G. ;Maher, C. ;Maia, J. ;Olds, T. ;Sarmiento, O. L. ;Standage, M. ;Tremblay, M. S. ;Tudor-Locke, C., International Study of Childhood Obesity, Lifestyle and the Environment (ISCOLE): Contributions to Understanding the Global Obesity Epidemic. Nutrients 2019, 11(4), 848. 2. Sabetghadam, A.; Grabowiecka-Nowak, A.; Kania, A.; Gugula, A.; Blasiak, E.; Blasiak, T.; Ma, S.; Gundlach, A. L.; Blasiak, A., Melanin-concentrating hormone and orexin systems in rat nucleus incertus: Dual innervation, bidirectional effects on neuron activity, and differential influences on arousal and feeding. Neuropharmacology 2018, 139, 238-256. 3. Kawata, Y. ;Okuda, S. ;Hotta, N. ;Igawa, H. ;Takahashi, M. ;Ikoma, M. ;Kasai, S. ;Ando, A. ;Satomi, Y. ;Nishida, M. ;Nakayama, M. ;Yamamoto, S. ;Nagisa, Y. ;Takekawa, S., A novel
15
and selective melanin-concentrating hormone receptor 1 antagonist ameliorates obesity and hepatic steatosis in diet-induced obese rodent models. Eur J Pharmacol 2017, 796, 45-53. 4. Lu, X.; Yang, H.; Chen, Y.; Li, Q.; He, S. Y.; Jiang, X.; Feng, F.; Qu, W.; Sun, H., The Development of Pharmacophore Modeling: Generation and Recent Applications in Drug Discovery. Curr Pharm Des 2018, 24(29), 3424-3439. 5. Dash, R. C. ;Ozen, Z. ;McCarthy, K. R. ;Chatterjee, N. ;Harris, C. A. ;Rizzo, A. A. ;Walker, G. C. ;Korzhnev, D. M. ;Hadden, M. K., Virtual Pharmacophore Screening Identifies SmallMolecule Inhibitors of the Rev1-CT/RIR Protein-Protein Interaction. ChemMedChem 2019, 14 (17), 1610-17. 6. Ulven, T.; Frimurer, T. M.; Receveur, J. M.; Little, P. B.; Rist, O.; Norregaard, P. K.; Hogberg, T., 6-Acylamino-2-aminoquinolines as potent melanin-concentrating hormone 1 receptor antagonists. Identification, structure-activity relationship, and investigation of binding mode. J Med Chem 2005, 48 (18), 5684-97. 7. Abu-Hammad, A.; Zalloum, W. A.; Zalloum, H.; Abu-Sheikha, G.; Taha, M. O., Homology modeling of MCH1 receptor and validation by docking/scoring and protein-aligned CoMFA. Eur J Med Chem 2009, 44 (6), 2583-96. 8. CATALYST 4.8 User Guide, A. I., San Diego, CA, USA, 2003. 9. Kurogi, Y.; Guner, O. F., Pharmacophore modeling and three-dimensional database searching for drug design using catalyst. Curr Med Chem 2001, 8 (9), 1035-55. 10. Guo, T.; Hunter, R. C.; Gu, H.; Rokosz, L. L.; Stauffer, T. M.; Hobbs, D. W., Discovery and SAR of 4-amino-2-biarylbutylurea MCH 1 receptor antagonists through solid-phase parallel synthesis. Bioorg Med Chem Lett 2005, 15 (16), 3691-5. 11. Su, J.; McKittrick, B. A.; Tang, H.; Czarniecki, M.; Greenlee, W. J.; Hawes, B. E.; O'Neill, K., Discovery of melanin-concentrating hormone receptor R1 antagonists using highthroughput synthesis. Bioorg Med Chem 2005, 13 (5), 1829-36. 12. Guo, T.; Shao, Y.; Qian, G.; Rokosz, L. L.; Stauffer, T. M.; Hunter, R. C.; Babu, S. D.; Gu, H.; Hobbs, D. W., Discovery and SAR of biaryl piperidine MCH1 receptor antagonists through solid-phase encoded combinatorial synthesis. Bioorg Med Chem Lett 2005, 15 (16), 3696-700. 13. Wu, W. L.; Burnett, D. A.; Caplen, M. A.; Domalski, M. S.; Bennett, C.; Greenlee, W. J.; Hawes, B. E.; O'Neill, K.; Weig, B.; Weston, D.; Spar, B.; Kowalski, T., Design and synthesis of orally efficacious benzimidazoles as melanin-concentrating hormone receptor 1 antagonists. Bioorg Med Chem Lett 2006, 16 (14), 3674-8. 14. Sasikumar, T. K.; Qiang, L.; Burnett, D. A.; Greenlee, W. J.; Hawes, B. E.; Kowalski, T. J.; O'Neill, K.; Spar, B. D.; Weig, B., Novel aminobenzimidazoles as selective MCH-R1 antagonists for the treatment of metabolic diseases. Bioorg Med Chem Lett 2006, 16 (20), 5427-31. 15. Meyers, K. M.; Mendez-Andino, J.; Colson, A. O.; Hu, X. E.; Wos, J. A.; Mitchell, M. C.; Hodge, K.; Howard, J.; Paris, J. L.; Dowty, M. E.; Obringer, C. M.; Reizes, O., Novel pyrazolopiperazinone- and pyrrolopiperazinone-based MCH-R1 antagonists. Bioorg Med Chem Lett 2007, 17 (3), 657-61. 16. Meyers, K. M.; Kim, N.; Mendez-Andino, J. L.; Hu, X. E.; Mumin, R. N.; Klopfenstein, S. R.; Wos, J. A.; Mitchell, M. C.; Paris, J. L.; Ackley, D. C.; Holbert, J. K.; Mittelstadt, S. W.; Reizes, O., Aminomethyl tetrahydronaphthalene biphenyl carboxamide MCH-R1 antagonists--Increasing selectivity over hERG. Bioorg Med Chem Lett 2007, 17 (3), 814-8.
16
17. Kamata, M.; Yamashita, T.; Imaeda, T.; Tanaka, T.; Masada, S.; Kamaura, M.; Kasai, S.; Hara, R.; Sasaki, S.; Takekawa, S.; Asami, A.; Kaisho, T.; Suzuki, N.; Ashina, S.; Ogino, H.; Nakano, Y.; Nagisa, Y.; Kato, K.; Ishihara, Y., Melanin-concentrating hormone receptor 1 antagonists. Synthesis and structure-activity relationships of novel 3(aminomethyl)quinolines. J Med Chem 2012, 55 (5), 2353-66. 18. Oost, T.; Heckel, A.; Kley, J. T.; Lehmann, T.; Müller, S.; Roth, G. J.; Rudolf, K.; Arndt, K.; Budzinski, R.; Lenter, M.; Lotz, R. R.; Maier, G. M.; Markert, M.; Thomas, L.; Stenkamp, D., Design, synthesis and evaluation of MCH receptor 1 antagonists--Part III: Discovery of pre-clinical development candidate BI 186908. Bioorg Med Chem Lett 2015, 25 (16), 3275-80. 19. Guner, O. F., Pharmacophore perception, development, and use in drug design. 1st ed.; International University Line: San Diego, CA, 2000. 20. Zhang, M.; Tamiya, J.; Nguyen, L.; Rowbottom, M. W.; Dyck, B.; Vickers, T. D.; Grey, J.; Schwarz, D. A.; Heise, C. E.; Haelewyn, J.; Mistry, M. S.; Goodfellow, V. S., Thienopyrimidinone bis-aminopyrrolidine ureas as potent melanin-concentrating hormone receptor-1 (MCH-R1) antagonists. Bioorg Med Chem Lett 2007, 17 (9), 2535-9. 21. Judd, A. S.; Souers, A. J.; Wodka, D.; Zhao, G.; Mulhern, M. M.; Iyengar, R. R.; Gao, J.; Lynch, J. K.; Freeman, J. C.; Falls, H. D.; Brodjian, S.; Dayton, B. D.; Reilly, R. M.; Gintant, G.; Limberis, J. T.; Mikhail, A.; Leitza, S. T.; Houseman, K. A.; Diaz, G.; Bush, E. N.; Shapiro, R.; Knourek-Segel, V.; Hernandez, L. E.; Marsh, K. C.; Sham, H. L.; Collins, C. A.; Kym, P. R., Identification of diamino chromone-2-carboxamides as MCHr1 antagonists with minimal hERG channel activity. Bioorg Med Chem Lett 2007, 17 (8), 236571. 22. Hudson, S.; Kiankarimi, M.; Rowbottom, M. W.; Vickers, T. D.; Wu, D.; Pontillo, J.; Ching, B.; Dwight, W.; Goodfellow, V. S.; Schwarz, D.; Heise, C. E.; Madan, A.; Wen, J.; Ban, W.; Wang, H.; Wade, W. S., Synthesis and structure-activity relationships of retro bisaminopyrrolidine urea (rAPU) derived small-molecule antagonists of the melaninconcentrating hormone receptor-1 (MCH-R1). Part 2. Bioorg Med Chem Lett 2006, 16 (18), 4922-30. 23. Su, J.; McKittrick, B. A.; Tang, H.; Burnett, D. A.; Clader, J. W.; Greenlee, W. J.; Hawes, B. E.; O'Neill, K.; Spar, B.; Weig, B.; Kowalski, T.; Sorota, S.; Li, C.; Liu, T., SAR study of bicyclo[4.1.0]heptanes as melanin-concentrating hormone receptor R1 antagonists: taming hERG. Bioorg Med Chem 2007, 15 (16), 5369-85. 24. Johansson, A., Evolution of physicochemical properties of melanin concentrating hormone receptor 1 (MCHr1) antagonists. Bioorg Med Chem Lett 2016, 26 (19), 4559-4564. 25. Daina, A.; Michielin, O.; Zoete, V., SwissADME: a free web tool to evaluate pharmacokinetics, drug-likeness and medicinal chemistry friendliness of small molecules. Sci Rep 2017, 7, 42717. 26. Helal, M. A.; Chittiboyina, A. G.; Avery, M. A., New insights into the binding mode of melanin concentrating hormone receptor-1 antagonists: homology modeling and explicit membrane molecular dynamics simulation study. J Chem Inf Model 2011, 51 (3), 635-46.
17
Highlights
Melanin Concentrating Hormone Receptor-1 is implicated in the regulation of food intake and energy homeostasis.
Two 3D pharmacophore models for MCHR-1 antagonists have been developed and validated.
The ZINC compound database was screened using the best models obtained.
Nine lead compounds showed more than 50% displacement of radiolabeled MCH at a 20 μM concentration.
The most active compounds were docked into a homology model of MCHR-1 to gain insights into their binding determinants.
18
19
20
21
22
Declaration of interests ☒ The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. ☐The authors declare the following financial interests/personal relationships which may be considered as potential competing interests:
23
Mohamed Helal October 1st, 2019
24