Comprehensive 3D-QSAR Model Predicts Binding Affinity of

Subscriber access provided by YORK UNIV

Pharmaceutical Modeling

Comprehensive 3D-QSAR Model Predicts Binding Affinity of Structurally Diverse Sigma 1 Receptor Ligands Youyi Peng, Hiep Dong, and William J. Welsh J. Chem. Inf. Model., Just Accepted Manuscript • DOI: 10.1021/acs.jcim.8b00521 • Publication Date (Web): 29 Nov 2018 Downloaded from http://pubs.acs.org on December 2, 2018

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

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

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

Journal of Chemical Information and Modeling

Comprehensive 3D-QSAR Model Predicts Binding Affinity of Structurally Diverse Sigma 1 Receptor Ligands Youyi Peng1*, Hiep Dong2, and William J Welsh1,3* 1: Biomedical Informatics Shared Resources, Rutgers Cancer Institute of New Jersey, Rutgers, The State University of New Jersey, 195 Little Albany Street, New Brunswick, NJ 08903, United States 2: Department of Medicinal Chemistry, Ernest Mario School of Pharmacy, Rutgers, The State University of New Jersey, 160 Frelinghuysen Road, Piscataway, New Jersey 08854, United States. 3: Department of Pharmacology, Robert Wood Johnson Medical School, Rutgers, The State University of New Jersey, 661 Hoes Lane West, Piscataway, NJ 08854, United States. *: Corresponding authors, [email protected], [email protected]

1 ACS Paragon Plus Environment

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

Page 2 of 33

Abstract The Sigma 1 Receptor (S1R) has attracted intense interest as a pharmaceutical target for various therapeutic indications, including the treatment of neuropathic pain and the potentiation of opioid analgesia. Efforts by drug developers to rationally design S1R antagonists have been spurred recently by the 2016 publication of the high-resolution X-ray crystal structure of the ligand-bound human S1R. Until now, however, the absence in the published literature of a single, large-scale, and comprehensive quantitative structure-activity relationship (QSAR) model that encompasses a structurally diverse collection of S1R ligands has impaired rapid progress. To our best knowledge, the present study represents the first report of a statistically robust and highly predictive 3D-QSAR model (R2 = 0.92, Q2 = 0.62, R2pred = 0.81) based on the X-ray crystal structure of human S1R and constructed from a pooled compilation of 180 S1R antagonists that encompass five structurally diverse chemical families investigated using identical experimental protocols. Best practices, as recommended by the Organization for Economic Cooperation and Development (OECD: http://www.oecd.org/), were adopted for pooling data from disparate sources and for QSAR model development and both internal and external model validation. The practical utility of the final 3D-QSAR model was tested by virtual screening of the DrugBank database of FDA approved drugs supplemented by eight reported S1R antagonists. Among the top ranked 40 DrugBank hits, four approved drugs which were previously unknown as S1R antagonists were tested using in vitro radiolabeled human S1R binding assays. Of these, two drugs (diphenhydramine and phenyltoloxamine) exhibited potent S1R binding affinity with Ki = 58 nM and 160 nM, respectively. As diphenhydramine is approved as an anti-allergic, and phenyltoloxamine as an analgesic and sedative, each of these compounds represents a viable starting point for a drug discovery campaign aimed at the development of novel S1R antagonists for a wide range of therapeutic indications.


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


Introduction The sigma receptors were initially identified in 1976 as an opioid receptor subtype and later as a phencyclidine binding receptor 1, 2. Now it is widely recognized that sigma receptors comprise a unique family distinct from opioid and phencyclidine receptors, and are classified into two subtypes: sigma 1 receptor (S1R) and sigma 2 receptor (S2R). The S1R has been cloned encoding a protein of 223 amino acids in diverse species including human, and has a molecular weight of 25.3 kDa expressed in both central and peripheral nervous systems

6, 7,

3-5.

It is highly

and has been identified as a ligand-gated

molecular chaperone protein within the endoplasmic reticulum (ER) and plasma membranes 8.

Recent research on the S1R has implicated its role in various pathological disorders. S1R knock-out (KO) mice have been shown to attenuate chemical-induced (e.g., formalin, capsaicin) and neuropathic (e.g., paclitaxel-induced) pain 9-12, and to potentiate opioid (e.g. morphine, oxycodone) analgesia but not its side effects (e.g., dependence, tolerance, constipation)

13-15.

Meanwhile, antagonists selective for the

S1R have been shown to elicit beneficial effects in neuropathic pain, the enhancement of opioid analgesia, and the mitigation of drug abuse including alcohol, cocaine and methamphetamine

16-18.

Emerging evidence also suggests that S1R antagonists inhibit tumor growth in prostate cancer xenograft models

19,

and induce autophagic degradation of programmed death-ligand 1 (PD-L1)

20.

On the other

hand, S1R agonists have been reported to play important roles in neuropsychiatric and neurodegenerative disorders including depression, epilepsy and Alzheimer’s disease 21-23.



A

Page 4 of 33

B Tyr103 Asp126

Leu105 Glu172 Leu95

Trp164

Met93

Phe107 Trp89

O

C

N

N N

O

S1RA (MR309/E-52862) Figure 1: A: Depiction of the human S1R in complex with PD144418 in the X-ray crystal structure (PDB ID: 5HK1). Viewing the receptor normal to the ER membrane reveals a trimeric overall architecture. Three protomers are shown in computer generated rendering, and PD144418 in the A chain is rendered as a green space-filled model. B: Binding pose of PD144418 (as green ball-and-stick model), showing the salt-bridge with Glu172 and hydrogen bond with Tyr103 in pink dotted line. Extensive hydrophobic contacts with other residues in the binding pocket are also shown and labeled accordingly. C: Structure of S1R antagonist S1RA. Despite the central role of the S1R in human physiology, the development of selective S1R ligands has been hampered by the lack of structural details until the recent publication of the X-ray crystallographic structure of the human S1R in complex with two ligands (PDB Ref Code 5HK1 and 5HK2) 24. The S1R crystal structure reveals a homotrimeric architecture with a single trans-membrane domain in each protomer (Figure 1A). The ligand-binding pocket in the S1R structure is deeply buried and completely occluded from solvent. The anionic sidechain of Glu172 forms a salt-bridge with the ligand’s cationic amine (Figure 1B), consistent with previous mutagenesis studies identifying this residue as essential for ligand binding

25.

Tyr103, an essential residue for ligand binding as shown by separate mutagenesis 4 ACS Paragon Plus Environment

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


experiments 26, interacts with the ligand through π-π stacking and forms a hydrogen bond with Glu172 to stabilize its orientation (Figure 1B). Other important residues in the binding site include Asp126, which in its protonated state forms a hydrogen bond with Glu172, and hydrophobic residues (Trp89, Met93, Leu95, Leu105, Phe107, and Trp164), which interact with hydrophobic or aromatic moieties in the bound ligand (Figure 1B). The detailed rendering of the ligand-receptor interactions afforded by the X-ray crystal structure is crucial for the structure-based design of novel S1R ligands.

S1RA (MR309/E-52862, Figure 1C), a S1R antagonist currently in Phase 2 clinical trials, has been reported to show encouraging results in peripheral neuropathy of different etiologies (including diabetic and chemotherapy-induced neuropathy) as well as in the potentiation of opioid analgesia without inducing adverse effects associated with opioid use such as tolerance

27.

Due to S1RA’s success in

preclinical and clinical studies and the availability of the human S1R crystal structure, S1R antagonists have recently attracted significant interest for myriad therapeutic indications. Quantitative structureactivity relationship (QSAR) models for S1R ligands have been developed in the past; however, these earlier models focused specific chemical families (e.g. benzylpiperazines, pentacycloundecylamines, penylacetamides, S1RA analogs, and spirocyclic compounds) and were based on homology models of the S1R

28-34,

which have since been demonstrated by the recently published S1R X-ray structures as

inaccurate with respect to the ligand-binding site and the receptor’s overall architecture

24, 29, 35.

As a

consequence, there remains an urgent need to develop predictive molecular models for chemically diverse S1R antagonists based on the S1R X-ray structures to facilitate the discovery of potent and selective S1R ligands using structure- and ligand-based molecular design approaches.

In the present study, pharmacophore-based 3D-QSAR models were developed for a pooled dataset of known S1R antagonists that encompasses five structurally diverse chemical families. The suitability of 5 ACS Paragon Plus Environment


Page 6 of 33

pooling data from different sources was assessed using best practices recommended by the Organization for Economic Cooperation and Development (OECD: http://www.oecd.org/) and validated by “leaveone-out” (LOO) cross-validation. The e-Pharmacophore modeling approach in the Schrödinger software (Schrödinger, LLC, New York, 2017) was employed using the X-ray crystal structure of human S1R for molecular alignment of all ligands for the QSAR study. Statistically significant and predictive 3DQSAR models for S1R antagonists were obtained, and further validated by the virtual screening against the DrugBank database

36.

By virtue of their robust performance and high predictive power, these

models can guide the identification of potent and selective S1R antagonists using virtual screening and rational de novo design techniques.

Materials and Methods All calculations were performed on an Intel Core i7 6700HQ 2.6 GHz processor with a memory of 8 GB RAM using Maestro 11.2 (Schrödinger, LLC, New York) or Molecular Operating Environment (MOE 2016.08, Chemical Computing Group, Montreal, QC, Canada). Data Collection Data retrieved from five separately published articles, using identical experimental binding assay protocols, were combined for our 3D-QSAR studies. Briefly, the competitive binding assays were performed against guinea pig brain membrane using [3H]-(+)pentazocine as the radiolabeled ligand. A total of 180 S1R antagonists were pooled for the development of 3D-QSAR models, yielding a dataset that spans >4 logarithmic (log) units in terms of binding affinity pKi (pKi = -logKi) and contains five distinct core structures (Table 1). The experimental biological activities of the compounds are well distributed: 38 weakly active compounds (pKi < 7.0), 95 moderately active compounds (7.0 ≤ pKi < 8.0), and 47 highly active compounds (pKi ≥ 8.0). 6 ACS Paragon Plus Environment

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


The dataset was divided into a modeling set (147 compounds) for model development and an external evaluation set (33 compounds) for model validation. The resulting external evaluation set, not employed in model building, contained 5 weakly active, 21 moderately active, and 7 highly active compounds. As is customary for QSAR studies, the modeling dataset was divided randomly between the training set and the testing set with the ratio of 4:1. This process was repeated 50 times with random selection of training/testing sets in order to assess the statistical variability and predictive power of the 3D-QSAR models. During the construction and validation of the 3D-QSAR models, normalization of some specific chemotypes (e.g., the quaternary amine ion, piperidine, and piperazine) was performed to avoid inconsistent results for all compounds in the modeling and external evaluation sets.

All compounds in the dataset were prepared using the LigPrep application implemented in Maestro 11.2, and protonated at the basic nitrogen atom at pH=7.4 for the QSAR modeling, since this feature is indispensable for all high affinity S1R ligands 24. Finally, energy minimization was conducted using the OPLS3 force field 37.

Conformer Generation Conformers of each ligand were generated by MacroModel implemented in Maestro, using the OPLS3 force field. Solvent effects were simulated using the Generalized-Born/Surface-Area (GB/SA) solvation model 38, with no cutoff value set for the non-bonded interactions. The Polak-Ribiere conjugate gradient (PRCG) method was then utilized for energy minimization with gradient convergence thresholds of 0.001 kJ/mol/Å and 2500 maximum iterations. Monte Carlo Multiple Minimum (MCMM) torsional sampling was employed to perform the conformational searches. Default values were set for the energy window to save structures and for the cutoff of maximum atom deviation to discard redundant 7 ACS Paragon Plus Environment


Page 8 of 33

conformers. All resulting conformers were aligned to the pharmacophore hypothesis generated in the Pharmacophore Hypothesis Generation section. The best fitting conformers of each compound were then selected and aligned for the 3D-QSAR model generation.

Principal Component Analysis (PCA) In order to identify potential outliers, which fall outside of the trend of the majority data, principal component analysis (PCA) was performed using MOE (ver. 2016.08). A total of 206 2D molecular descriptors were computed for each molecule in the entire database, from which the principal components (PCs) were calculated. Then, a 3D graphical plot was built with respect to the first three principal components (PC1, PC2 and PC3). A molecule is considered as an outlier when it is located distant from the majority of data in the 3D graphical plot. Using this approach, no outliers were identified.

Pharmacophore Hypothesis Generation The e-Pharmacophore method in Phase implemented in Maestro was used with all default settings to develop the pharmacophore hypothesis based on key interactions and the binding site volume information from the X-ray crystal structure of the S1R in complex with its antagonist PD144418 (PDB Ref Code 5HK1), downloaded from the PDB (www.rcsb.org). First, the Protein Preparation wizard in Maestro was first employed to prepare the protein structure, during which the hydrogen bond network was optimized, then the protein structure was energy-minimized using the OPLS3 force field to eliminate steric clashes. All water molecules in the crystal structure were removed prior to the calculations.


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


Second, the co-crystal ligand PD144418 was extracted from the X-ray crystal structure and then redocked into the binding pocket of S1R using Glide XP

39.

Energy-based descriptors from the Glide

docking were utilized to generate and rank the pharmacophore features. Phase contains six chemical features: hydrogen-bond donor (D), hydrogen-bond acceptor (A), aromatic ring (R), positively charged group (P), negatively charged group (N), and hydrophobic group (H), and additionally excluded volumes, which represent regions of space that are precluded to the ligand inside the receptor’s binding when it is aligned to the pharmacophore. During the process of the e-Pharmacophore generation, PD144418 was selected as the reference ligand for the Glide docking-scoring procedure. In our study, the final pharmacophore hypothesis comprised the excluded volume together with three important features based on their contribution to binding energy: aromatic ring group (R), positively charged group (P), and hydrogen-bond acceptor group (A).

3D-QSAR Model Development The modeling set of 147 compounds was used for the development of the 3D-QSAR models employing the standard procedure implemented in the Phase application. This atom-based QSAR approach takes into account all atoms of the compounds, not merely the pharmacophoric features

40.

The partial least

squares (PLS) technique was employed to generate a linear relationship that correlates changes in the computed molecular descriptors with changes in the corresponding experimental values of the binding affinity (pKi) for compounds in the dataset. PLS converts the original molecular descriptors to so-called latent variables, similar to principal components (PCs), that consist of linear combinations of the original independent variables. To assess the internal predictive ability of the 3D-QSAR models, LOO cross-validation was employed during which each compound was excluded one at a time, and its activity was predicted by the model 9 ACS Paragon Plus Environment


Page 10 of 33

constructed from the remaining compounds in the dataset. Cross-validation determines the optimum number of PCs, corresponding to the smallest error of prediction and the highest cross-validated Q2 (or R2cv). PLS analysis was repeated for the entire modeling set using the optimum number of PCs to generate a final 3D-QSAR model by keeping the grid spacing at 0.6 Å and eliminating variables with tvalue < 2.0 to reduce “noise” and to improve efficiency. All 3D-QSAR models are represented as color contour maps to enable visualization of those molecular property features that significantly contribute to the model.

3D-QSAR Model Based Virtual Screening The predictive power of the 3D-QSAR models was assessed using an external evaluation dataset as described earlier. In order to further assess the predictability of our 3D-QSAR models, virtual screening was performed against the DrugBank database

36

downloaded from https://www.drugbank.ca which

contained 2334 drugs approved by the US Food and Drug Administration (FDA). Peptide-like and inorganic molecules, together with drugs violating Lipinski’s Rule of Five (RO5)

41,

were eliminated

from selection, resulting in a filtered DrugBank database of 1935 drugs for the current study. Moreover, eight known S1R antagonists (S1RA, DuP 734, MS-377, E-100, E-5482, BD-1063, BD-1008 and BMS181100) 13, 42-46, which were not included in the development of our 3D-QSAR model, were added to the DrugBank database, yielding a total of 1943 compounds for this validation dataset. All compounds in this virtual screening study were prepared and screened against the pharmacophore hypothesis using the same protocol as described above for the 3D-QSAR model development. The best fitting conformers of each compound were submitted to 3D-QSAR models for prediction of their S1R binding affinity (pKi). The top-ranked 40 hits were explored to assess the performance of the 3D-QSAR model for virtual screening.


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


The experimental S1R binding affinity values of the top 40 hits from the virtual screening were retrieved from the ChEMBL database (https://www.ebi.ac.uk/chembl/)

47

and DrugMatrix Pharmacology Data

(https://ntp.niehs.nih.gov/drugmatrix/index.html) 48.

In Vitro Radiolabeled Binding Assay The radiolabeled binding assay of the S1R was performed by Eurofins Panlabs Discovery Services (Taipei, Taiwan) using human jurkat cell membranes according to the protocol previously reported

49.

The binding assay was conducted in duplicate on membrane preparations that had been re-suspended in 50 mM Tris-HCl at pH=8.0 using [3H]-pentazocine (15.0 nM) as the radioligand. Following 2-hr incubation at 37 oC, the binding assay was terminated by addition of cold buffer. The mixture was then filtered through Whatman GF/B filters and washed with cold buffer. Radioactivity was identified using the TopCount NTX liquid scintillation counter (PerkinElmer, Waltham, MA). Non-specific binding for the S1R was measured in the presence of 10 µM unlabeled haloperidol. Receptor binding data were analyzed by nonlinear regression of saturation and competition curves using the GraphPad Prism 7.0 software (GraphPad Software, La Jolla, CA).

Results and Discussion Data Collection and Curation OECD best practices were followed to develop statistically robust and predictive 3D-QSAR models in the current study

50.

In order to achieve statistically significant 3D-QSAR models, several heuristic

rules were adopted in selecting the dataset: (1) the minimum range of bioactivity should be three log units, (2) the biological activities of the molecules should be distributed evenly throughout the range, and (3) the training dataset should contain at least 40 compounds to develop a QSAR model that can predict continuous response variables (e.g., binding affinity)

51.


Ideally, all compounds should come


Page 12 of 33

from a single source so as to minimize the inevitable variation in pharmacological data measured by different laboratories 52. When this is not possible, as is the case with the subject S1R antagonists, extra precautions should be taken to validate pooling data from different sources, which has been shown to be a feasible strategy by our previous QSAR studies 53, 54.

One mandatory requirement for pooling such biological data is that they are derived using identical experimental protocols. We retrieved biological data from five independent publications, each of which reported results for 21-51 S1R antagonists using the same radiolabeled ligand [3H]-(+) pentazocine in the competitive binding assay. Chiral compounds without specified stereoisomers and replicate structures were excluded from our dataset. A dataset of 180 S1R antagonists was compiled for development of the QSAR models, with >4 log units in the pKi values and five distinct core structures (Table 1). The experimental biological activities (pKi) of the compounds are well distributed: 38 weakly active compounds, 95 moderately active compounds, and 47 highly active compounds. Table 1: Structures pooled from separate publications for the present 3D-QSAR studies Structure X

O

Z

O N O O

N Y

N

Z

Core No. No. of Compounds Range of Exp. pKi Ref Y

1

37

6.40-9.52

55

2

37

6.24-8.85

56

3

21

6.07-8.95

57

4

34

6.72-8.91

58

X


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


5

51

6.01-9.01

59

Specific substituent moieties X, Y, Z, R1 and R2 are shown in Table S1 (Supporting Information). A fundamental assumption of all QSAR studies is that structurally similar compounds possess similar activities or properties. However, it is not uncommon that small structural changes in some compounds can translate to large changes in activities and properties, in violation of this principle. Such compounds are classified as structural or activity outliers, and should be avoided in any QSAR modeling study

51.

To address this issue, we employed PCA on the 2D molecular descriptors computed using MOE for the entire dataset. A molecule is considered an outlier if it is located distant from the majority of data in the 3D graphical plot of the first three principal components (PC1, PC2 and PC3). Fortunately, this PCA confirmed the absence of any potential outliers or “cliffs” in the dataset (Figure S1 in Supporting Information). The entire dataset of 180 compounds was then separated into a modeling set (147 compounds) for model development and an external evaluation set (33 compounds) for model validation. The evaluation set compounds were randomly chosen with some bias toward ensuring representation from the full range of biological data in the modeling set: 5 weakly active, 21 moderately active, and 7 highly active compounds. The size of the evaluation set was about 18% of the entire dataset (180 compounds), which conforms with the recommended range of 15%-20% by Tropsha 51, and was employed to validate the predictive power of the 3D-QSAR models.

Generation of Structure-Based e-Pharmacophore Models A major challenge when developing 3D-QSAR models is the molecular alignment of the training and testing set compounds, along with the conformation and orientation of each ligand selected for the 13 ACS Paragon Plus Environment


Page 14 of 33

modeling. Recently, the e-Pharmacophore approach (implemented in Maestro/Phase) based molecular alignment was utilized to develop predictive 3D-QSAR models for lead optimization and virtual screening of novel ligands 60-63. Similarly, we applied the e-Pharmacophore approach in the present 3DQSAR studies to achieve better alignment. The X-ray crystal structure of human S1R and its known antagonist PD144418 (PDB ID: 5HK1) was used to generate the pharmacophore hypothesis. The native ligand PD144418 was chosen as the reference molecule for the energy-based calculations using Glide, from which three pharmacophoric features were selected (Figure 2): aromatic ring (R), positively charged group (P), and hydrogen bond acceptor (A). The R and P features have been shown by visual analysis of the X-ray crystal structures and experimental mutagenesis studies to play key roles in protein-ligand recognition and binding 25, 26. Specifically, the feature P involves a salt-bridge interaction with Glu172, while the feature R engages in π-π stacking with the sidechain of Tyr103, and in hydrophobic interactions with residues Met93, Leu95 and Leu105 in the S1R binding pocket. Some inconsistency was observed for the hydrogen-bond acceptor feature. Based on the X-ray crystal structure of S1R, no hydrogen-bond donors were identified in the receptor-binding pocket to pair with the acceptor. However, visual examination of all ligands in the current dataset found an electronegative atom at this position. Previous pharmacophore modeling studies also identified the hydrogen bond acceptor as a crucial feature for potent S1R antagonists 64, 65, although other studies regarded this feature as insignificant 66-68. We speculate that the hydrogen bond acceptor was likely introduced in the ligands to confer improved physico-chemical properties or to facilitate chemical synthesis, not on the grounds of satisfying a specific pharmacophoric feature. Nevertheless, the hydrogen-bond acceptor feature (A) was retained in the current modeling study for proper alignment of all compounds.


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


Asp126

Glu172

Tyr103 P

R

A

Figure 2: Left: X-ray crystal structure of the human S1R with PD144418 (PDB ID: 5HK1), showing the ligand (green) surrounded by the surface of binding pocket (light blue), and key residues (dark gray). The magenta and red dash lines depict two key interactions between the S1R and its ligand, namely the salt-bridge and π-π stacking, respectively. Right: Four features included in the e-Pharmacophore model: positive charged group (P), hydrogen bond acceptor (A), aromatic ring (R), and excluded volume (light blue). The native ligand PD144418 (green) and the most potent ligand (34, magenta) are aligned to the pharmacophore model.

In addition to ligand-based pharmacophoric features, structure-based pharmacophore models may include an excluded volume, which defines a specific region inside the receptor-binding pocket that is occupied by high-affinity ligands. Notably, the excluded volume should be implemented in pharmacophore models only if there is evidence of the rigidity of the binding pocket, which is the case here for the S1R

24.

As shown by the X-ray crystal structure, the ligand-binding pocket of the S1R is

completely buried and occluded from the solvent. Therefore, an excluded volume feature was constructed from the size of the S1R binding pocket and was added to the pharmacophore model. Altogether, four features were selected for the pharmacophore hypothesis, which was utilized for the alignment of all S1R ligands in the development of the 3D-QSAR models 40.

Generation of Atom-based 3D-QSAR Models The entire dataset of 180 compounds, pooled from five separate published studies, was divided into a modeling dataset of 147 compounds for model development and an external evaluation dataset of 33 15 ACS Paragon Plus Environment


Page 16 of 33

compounds for assessing the model’s predictability. LOO cross-validation was employed to assess the internal predictive ability of the 3D-QSAR models. Exploratory 3D-QSAR studies on the modeling dataset (results not shown) suggested that four PCs would yield optimal statistics in terms of R2 (squared correlation coefficient of the regression model, a measure of internal consistency) and Q2 (squared correlation coefficient of the leave-one-out cross validation, a measure of internal predictability). In order to achieve statistical significance and to reduce the risk of over-fitting, four PCs in the PLS model were chosen for the subsequent studies. To fully evaluate the statistical variability and predictive power of the 3D-QSAR models, the entire modeling dataset of 147 compounds was divided randomly between the training set and the testing set with the ratio of 4:1. This process was repeated 50 times in order to avoid the coincidence. The results for the 50 separate 3D-QSAR models, summarized in Table 2, demonstrated excellent statistical quality and stability, robust cross-validation and internal predictability with R2 = 0.90-0.93, Q2 = 0.57-0.60, and R2pred = 0.63-0.70. R2pred, defined analogously to Q2, is used to evaluate the overall performance of a model by comparing the accuracy of a series of predictions with the experimental target property values for a testing set. Table 2: Statistics of the Atom-based 3D-QSAR Models with Four PCs Statistical Parameters Range of 50 Models Final Model* Number of molecules in the training set 117 147 Number of molecules in the testing set 30 0 Principal Components (PCs) 4 4 2 R 0.90 - 0.93 0.92 Q2 (LOO) 0.57 – 0.60 0.62 SD 0.19 – 0.36 0.24 F value 272 – 349 385 RMSE 0.48 – 0.41 0.29 𝑅2𝑝𝑟𝑒𝑑 0.63 – 0.70 0.81 2 2 R : Squared correlation coefficient of the regression model. Q : Cross-validated correlation coefficient. SD: Standard deviation. F: Variance ratio. RMSE: Root-mean-square error. R2pred: Correlation coefficient between actual and predicted activities (pKi) of the test set compounds. *: The final 3DQSAR model was built for the entire modeling dataset, i.e., without partitioning of the training and testing sets. Table 3: Statistics of Atom-based 3D-QSAR Models in External Validation Studies 16 ACS Paragon Plus Environment

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


Statistical Parameters Calculated Value Criteria Q2 0.62 Q2 > 0.5 Rpred 0.90 Close to 1 𝐑𝟐𝐩𝐫𝐞𝐝 R2pred > 0.6 0.81 k 1.00 0.85 ≤ k ≤ 1.15 k’ 0.97 0.85 ≤ k’ ≤ 1.15 𝐑𝟐𝐨 0.79 Close to R2 𝐑′𝐨𝟐 0.81 Close to R2 𝟐 𝟐 𝟐 (𝐑𝐩𝐫𝐞𝐝 - 𝐑𝐨)/ 𝐑𝐩𝐫𝐞𝐝 0.03 < 0.1 (𝐑𝟐𝐩𝐫𝐞𝐝 - 𝐑′𝟐𝐨)/ 𝐑𝟐𝐩𝐫𝐞𝐝 0.00 < 0.1 2 2 Q : Cross-validated coefficient. R (or Rpred): Correlation coefficient between experimental and predicted activities. k: Slope for the regression line of experimental versus predicted activities through the origin. k’: Slope for the regression line of predicted and experimental activities through the origin. R2o: Correlation coefficient between experimental and predicted activities through the origin. R′o2: Correlation coefficient between predicted and experimental activities through the origin. A final 3D-QSAR model with four PCs was built for the entire modeling dataset to yield R2 = 0.92 and Q2 = 0.62, with an acceptably low standard deviation (SD) of 0.24, high F value (ratio of variance explained by models and variance due to error in regression) of 385, and low root-mean-square error (RMSE) of 0.29 (Table 2). These statistics corroborate that the 3D-QSAR model exhibits excellent consistency and good internal predictability. The strong correlation between QSAR-predicted and experimental pKi values for the S1R antagonists is evident from the plot (Figure 3) and from the tabulated results (Table S1 in Supporting Information). The residuals (differences) between corresponding values of experimental and predicted binding affinity are all = 6) from the literature

42, 43, 45, 46, 71, 72,

the ChEMBL database

47

Page 22 of 33

and DrugMatrix Pharmacology

Data 48. Seven of the eight known S1R ligands are on the top 40 list, with only one known ligand (BD1008) failing to meet this threshold. Conversely, only three (scopolamine, phenoxybenzamine and dolasetron) out of the top 40 drugs were predicted as potent S1R ligands (pKi > 8), but are classified as inactives according to the binding affinity data (pKi < 5) from the DrugMatrix database. Notably, dolasetron and phenoxybenzamine are racemic compounds, and scopolamine contains a bridged polycyclic. Both racemic and bridged compounds fall outside of the applicability domains of the present 3Q-QSAR models, which may explain their false positive prediction as S1R ligands. Table 4: Four Drugs from the DrugBank Database for Human S1R Binding Assays* Generic Name Structure Inh% at 1μM N O

Phenyltoloxamine

Diphenhydramine

66%

70%

N

O

10%

Amisulpride

O

Metoclopramide

Cl H 2N

N H

N

34%

O

*: The binding assay was performed by Eurofins Panlabs Discovery Services in human jurkat cell membranes using [3H]-pentazocine as the radioligand, as previously described 49. Interestingly, twelve of the top 40 drugs have not been tested previously in the S1R binding assay. Based on their commercial availability and structural diversity, four of these drugs (phenyltoloxamine, diphenhydramine, metoclopramide, and amisulpride) were acquired for initial biological evaluation of 22 ACS Paragon Plus Environment

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


their human S1R binding affinity at a concentration of 1 μM (Table 4). The S1R binding assay was conducted in duplicate on human jurkat cell membrane preparations using [3H]-pentazocine (15.0 nM) as the radioligand according to the protocol previously reported49. Phenyltoloxamine and diphenhydramine exhibited 66% and 70% binding affinity to the human S1R, respectively. On the other hand, metoclopramide and amisulpride showed 34% and 10% binding affinity to the human S1R that fall below our selected cut-off value (i.e., 50% inhibition at 1 μM). Notably amisulpride, as a racemic compound, falls outside of the applicability domains of the present model, which may explain its false positive prediction as a S1R ligand. A follow-up assay was then run for phenyltoloxamine and diphenhydramine to determine their S1R binding affinity (Ki). Remarkably, diphenhydramine and phenyltoloxamine demonstrated potent binding affinity for the human S1R with Ki = 58 nM and Ki = 160 nM, respectively (Figure 6).

Ki = 58 nM

Ki = 160 nM

Figure 6: Dose-response curves of diphenylhydramine (left, blue circles) and phenyltoloxamine (right, blue circles) in human S1R radiolabeled binding assays. Haloperidol is in red squares, and Ki values are indicated accordingly. Diphenhydramine is an antihistamine drug, mainly indicated for the treatment of allergies 36. Its primary target is the histamine (H1) receptor (Ki = 15 nM), but is known to bind other targets including the muscarinic acetylcholine receptors, the serotonin 2 (5-HT2) receptors, and the norepinephrine 23 ACS Paragon Plus Environment


transporter (Ki in the range of 50 nM to 3.5 μM Table 5)

Page 24 of 33

47, 48.

Diphenhydramine has been reported to

73.

Similarly, phenyltoloxamine is also an

potentiate the analgesic effects of opioids like morphine

36

by

potentiating the inherent analgesic effects of acetaminophen as well as codeine and its derivatives

74.

antihistamine drug with sedative and analgesic effects, and also acts as an adjuvant analgesic

Phenyltoloxamine binds to its primary target histamine 1 (H1) receptor (Ki = 12.6 nM) as well as offtargets such as the dopamine D1-D5 receptors (Ki in the range of 340 nM to 10 μM, Table 6)

47, 48.

Nevertheless, the mechanism(s) by which phenyltoloxamine and diphenhydramine potentiate the antinociceptive effects of known analgesics remains elusive. The finding reported here that diphenhydramine and phenyltoloamine are high-affinity S1R ligands (Ki = 58 nM and Ki = 160 nM, respectively), may help the biomedical community to elucidate the mechanisms associated with their direct and indirect (potentiating) analgesic effects. Table 5: Pharmacological Activities of Diphenhydramine Target Name

Ki (nM)

Sigma 1 receptor (S1R)

58

Histamine H1 receptor

15

Muscarinic acetylcholine receptor M4

52


83


116


137

Serotonin 2a (5-HT2a) receptor

370


373

Serotonin 2c (5-HT2c) receptor

513

Serotonin 2b (5-HT2b) receptor

711

Norepinephrine transporter

3513

Table 6: Pharmacological Activities of Phenyltoloxamine Target Name

Ki (nM)

Sigma 1 receptor (S1R)

160


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


Histamine H1 receptor

12.6

Dopamine D1 receptor

343


414


2740


2765


10000

Conclusions Pharmacophore-based 3D-QSAR models were constructed to develop predictive models for virtual screening and rational design of novel S1R ligands and, more generally, to gain insights into the structure-activity relationships of S1R antagonists. Best practices recommended by Tropsha and the OECD were followed in order to develop statistically robust and predictive 3D-QSAR models. In the present study, a large and diverse data set of S1R antagonists was created by pooling ligand information from independent sources. Heuristic rules of thumb were adopted when compiling the dataset to obtain an optimal range and even distribution of bioactivities (pKi), and to maximize the number of ligands. PCA was conducted to eliminate structural or activity outliers from the data set assembled for the QSAR studies. A pharmacophore model with four features (R, P, A and excluded volume) was developed based on the S1R crystal structures and employed as the template for the molecular alignment of all ligands. PLS analysis, together with LOO cross-validation, was utilized for model development and validation. Atom-based 3D-QSAR models were constructed by randomly dividing the modeling set into a training and testing set with the ratio of 4:1, to yield a final model for the S1R antagonists with excellent consistency (R2 = 0.92) and internal predictability (Q2 = 0.62). An external evaluation dataset was then selected to evaluate the predictability of the 3D-QSAR models, thus confirming its predictive ability ( 25 ACS Paragon Plus Environment


Page 26 of 33

R2pred = 0.81) and consistently robust model-based statistics (including k, k’, R2o and R′o2) in accordance with all such criteria set by the OECD.

3D-QSAR model based virtual screening was carried out against the DrugBank database, in order to further evaluate the predictive power of our 3D-QSAR models. Twenty-five out of the top 40 drugs predicted by our 3D-QSAR models as S1R ligands were confirmed to be true S1R ligands. Only three out of the top 40 drugs were found to be false positives. Two drugs (phenyltoloxamine, diphenhydramine), which to our knowledge have not been tested for S1R binding affinity before, were found to exhibit potent binding affinity to human S1R (Ki = 58 nM and Ki = 160 nM, respectively) in radiolabeled receptor binding assays. It is hoped that the discovery reported here that both phenyltoloxamine and diphenhydramine, while quite promiscuous (Tables 5 and 6), also exhibit high S1R binding affinity will help interested scientists and clinicians to understand the mechanisms for their therapeutic uses and side effects, especially their ability to potentiate the analgesic effects of opioid analgesics, and possibly to repurpose these drugs for therapeutic indications associated with their S1R activity.

The present report represents the first successful attempt to develop a comprehensive 3D-QSAR model, based on the recently published X-ray structure of human S1R, sourced by pooling and curating a large assemblage of structurally diverse S1R antagonists, which should prove useful for the identification of new drug leads and for prediction of their S1R binding affinity (pKi) prior to the resource-demanding tasks of chemical synthesis and experimental biological evaluation. Moreover, these 3D-QSAR models might be useful as a filter to exclude molecules that exhibit off-target activity at the S1R in drug discovery campaigns. As a final admonition, it should be emphasized that these 3D-QSAR models are


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


statistically valid only within their applicability domains with respect to their range of binding affinities (pKi = 6~10) and the chemical space encompassed by the S1R ligands in the present study.

Acknowledgments This research has been supported in part with a grant from the New Jersey Health Foundation, by the Rutgers TechAdvance/TechXpress fund, and by the Biomedical Informatics Shared Resource of the Rutgers Cancer Institute of New Jersey (P30CA072720). We gratefully acknowledge access to the High Performance Computing facilities and support of the computational STEM and bioinformatics scientists (especially Dr. Vladyslav Kholodovych) at the Rutgers Office of Advanced Research Computing (OARC, URL: http://oarc.rutgers.edu). We also acknowledge the computational resources made possible through the access to the Perceval Linux cluster operated by OARC under NIH 1S10OD012346-01A1. Notes The authors declare no competing financial interest.

Supporting Information Available: Structure and activity data of all S1R antagonists used for the present 3D-QSAR studies, the principal component analysis plot of S1R Antagonists, and structure and activity data of top 40 hits from the screening of the DrugBank Database. This material is available free of charge via the Internet at http://pubs.acs.org.

References 1. Martin, W. R.; Eades, C. G.; Thompson, J. A.; Huppler, R. E.; Gilbert, P. E., The Effects of Morphine- and Nalorphine- Like Drugs in the Nondependent and Morphine-Dependent Chronic Spinal Dog. J. Pharmacol. Exp. Ther. 1976, 197, 517-32.



Page 28 of 33

2. Khazan, N.; Young, G. A.; El-Fakany, E. E.; Hong, O.; Calligaro, D., Sigma Receptors Mediated the Psychotomimetic Effects of N-Allylnormetazocine (SKF-10,047), but Not Its Opioid Agonistic-Antagonistic Properties. Neuropharmacology 1984, 23, 983-7. 3. Hanner, M.; Moebius, F. F.; Flandorfer, A.; Knaus, H. G.; Striessnig, J.; Kempner, E.; Glossmann, H., Purification, Molecular Cloning, and Expression of the Mammalian Sigma1-Binding Site. Proc. Natl. Acad. Sci. U. S. A. 1996, 93, 8072-7. 4. Seth, P.; Fei, Y. J.; Li, H. W.; Huang, W.; Leibach, F. H.; Ganapathy, V., Cloning and Functional Characterization of a Sigma Receptor from Rat Brain. J. Neurochem. 1998, 70, 922-31. 5. Pan, Y. X.; Mei, J.; Xu, J.; Wan, B. L.; Zuckerman, A.; Pasternak, G. W., Cloning and Characterization of a Mouse Sigma1 Receptor. J. Neurochem. 1998, 70, 2279-85. 6. Alonso, G.; Phan, V.; Guillemain, I.; Saunier, M.; Legrand, A.; Anoal, M.; Maurice, T., Immunocytochemical Localization of the Sigma(1) Receptor in the Adult Rat Central Nervous System. Neuroscience 2000, 97, 155-70. 7. Bangaru, M. L.; Weihrauch, D.; Tang, Q. B.; Zoga, V.; Hogan, Q.; Wu, H. E., Sigma-1 Receptor Expression in Sensory Neurons and the Effect of Painful Peripheral Nerve Injury. Mol. Pain 2013, 9, 47. 8. Hayashi, T.; Su, T. P., Sigma-1 Receptor Chaperones at the ER-Mitochondrion Interface Regulate Ca(2+) Signaling and Cell Survival. Cell 2007, 131, 596-610. 9. Cendan, C. M.; Pujalte, J. M.; Portillo-Salido, E.; Montoliu, L.; Baeyens, J. M., Formalin-Induced Pain Is Reduced in Sigma(1) Receptor Knockout Mice. Eur. J. Pharmacol. 2005, 511, 73-4. 10. Entrena, J. M.; Cobos, E. J.; Nieto, F. R.; Cendan, C. M.; Gris, G.; Del Pozo, E.; Zamanillo, D.; Baeyens, J. M., Sigma-1 Receptors Are Essential for Capsaicin-Induced Mechanical Hypersensitivity: Studies with Selective Sigma-1 Ligands and Sigma-1 Knockout Mice. Pain 2009, 143, 252-61. 11. Nieto, F. R.; Cendan, C. M.; Sanchez-Fernandez, C.; Cobos, E. J.; Entrena, J. M.; Tejada, M. A.; Zamanillo, D.; Vela, J. M.; Baeyens, J. M., Role of Sigma-1 Receptors in Paclitaxel-Induced Neuropathic Pain in Mice. J. Pain 2012, 13, 1107-21. 12. Nieto, F. R.; Cendan, C. M.; Canizares, F. J.; Cubero, M. A.; Vela, J. M.; Fernandez-Segura, E.; Baeyens, J. M., Genetic Inactivation and Pharmacological Blockade of Sigma-1 Receptors Prevent Paclitaxel-Induced Sensory-Nerve Mitochondrial Abnormalities and Neuropathic Pain in Mice. Mol. Pain 2014, 10, 11. 13. Sanchez-Fernandez, C.; Montilla-Garcia, A.; Gonzalez-Cano, R.; Nieto, F. R.; Romero, L.; Artacho-Cordon, A.; Montes, R.; Fernandez-Pastor, B.; Merlos, M.; Baeyens, J. M.; Entrena, J. M.; Cobos, E. J., Modulation of Peripheral Mu-Opioid Analgesia by Sigma1 Receptors. J. Pharmacol. Exp. Ther. 2014, 348, 32-45. 14. Sanchez-Fernandez, C.; Nieto, F. R.; Gonzalez-Cano, R.; Artacho-Cordon, A.; Romero, L.; Montilla-Garcia, A.; Zamanillo, D.; Baeyens, J. M.; Entrena, J. M.; Cobos, E. J., Potentiation of Morphine-Induced Mechanical Antinociception by Sigma(1) Receptor Inhibition: Role of Peripheral Sigma(1) Receptors. Neuropharmacology 2013, 70, 348-58. 15. Vidal-Torres, A.; de la Puente, B.; Rocasalbas, M.; Tourino, C.; Bura, S. A.; Fernandez-Pastor, B.; Romero, L.; Codony, X.; Zamanillo, D.; Buschmann, H.; Merlos, M.; Baeyens, J. M.; Maldonado, R.; Vela, J. M., Sigma-1 Receptor Antagonism as Opioid Adjuvant Strategy: Enhancement of Opioid Antinociception without Increasing Adverse Effects. Eur. J. Pharmacol. 2013, 711, 63-72. 16. Almansa, C.; Vela, J. M., Selective Sigma-1 Receptor Antagonists for the Treatment of Pain. Future Med. Chem. 2014, 6, 1179-99. 17. Arena, E.; Dichiara, M.; Floresta, G.; Parenti, C.; Marrazzo, A.; Pittala, V.; Amata, E.; Prezzavento, O., Novel Sigma-1 Receptor Antagonists: From Opioids to Small Molecules: What Is New? Future Med. Chem. 2018, 10, 231-256. 18. Navarro, G.; Moreno, E.; Aymerich, M.; Marcellino, D.; McCormick, P. J.; Mallol, J.; Cortes, A.; Casado, V.; Canela, E. I.; Ortiz, J.; Fuxe, K.; Lluis, C.; Ferre, S.; Franco, R., Direct Involvement of Sigma-1 Receptors in the Dopamine D1 Receptor-Mediated Effects of Cocaine. Proc. Natl. Acad. Sci. U. S. A. 2010, 107, 18676-81.


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


19. Thomas, J. D.; Longen, C. G.; Oyer, H. M.; Chen, N.; Maher, C. M.; Salvino, J. M.; Kania, B.; Anderson, K. N.; Ostrander, W. F.; Knudsen, K. E.; Kim, F. J., Sigma1 Targeting to Suppress Aberrant Androgen Receptor Signaling in Prostate Cancer. Cancer Res. 2017, 77, 2439-2452. 20. Maher, C. M.; Thomas, J. D.; Haas, D. A.; Longen, C. G.; Oyer, H. M.; Tong, J. Y.; Kim, F. J., Small-Molecule Sigma1 Modulator Induces Autophagic Degradation of PD-L1. Mol. Cancer Res. 2018, 16, 243-255. 21. Kulkarni, S. K.; Dhir, A., Sigma-1 Receptors in Major Depression and Anxiety. Expert Rev. Neurother. 2009, 9, 1021-34. 22. Jin, J. L.; Fang, M.; Zhao, Y. X.; Liu, X. Y., Roles of Sigma-1 Receptors in Alzheimer's Disease. Int. J. Clin. Exp. Med. 2015, 8, 4808-20. 23. Nguyen, L.; Lucke-Wold, B. P.; Mookerjee, S. A.; Cavendish, J. Z.; Robson, M. J.; Scandinaro, A. L.; Matsumoto, R. R., Role of Sigma-1 Receptors in Neurodegenerative Diseases. J. Pharmacol. Sci. 2015, 127, 17-29. 24. Schmidt, H. R.; Zheng, S.; Gurpinar, E.; Koehl, A.; Manglik, A.; Kruse, A. C., Crystal Structure of the Human σ1 Receptor. Nature 2016, 532, 527. 25. Seth, P.; Ganapathy, M. E.; Conway, S. J.; Bridges, C. D.; Smith, S. B.; Casellas, P.; Ganapathy, V., Expression Pattern of the Type 1 Sigma Receptor in the Brain and Identity of Critical Anionic Amino Acid Residues in the Ligand-Binding Domain of the Receptor. Biochim. Biophys. Acta 2001, 1540, 59-67. 26. Yamamoto, H.; Miura, R.; Yamamoto, T.; Shinohara, K.; Watanabe, M.; Okuyama, S.; Nakazato, A.; Nukada, T., Amino Acid Residues in the Transmembrane Domain of the Type 1 Sigma Receptor Critical for Ligand Binding. FEBS Lett. 1999, 445, 19-22. 27. Cebrecos, J.; Galvez, R.; Albesa, N.; MunozBlanco, J.; Aguilar, J.; Rojals, V. M.; Casals, S.; Sust, M.; Vaqué, A.; Morte, A.; Gascón, N.; PlataSalamán, C. E52862, a First in Class Sigma₁ Receptor Antagonist, in Chronic Postsurgical Neuropathic Pain: An Exploratory Phase Ii Clinical Trial. In 16th World Congree on Pain, Yokohama, Japan, 2016; Yokohama, Japan, 2016. 28. Nahas, R. I.; Lever, J. R.; Lever, S. Z., Synthesis and Structure-Activity Relationships of N-(3-Phenylpropyl)N'-Benzylpiperazines: Potent Ligands for Sigma1 and Sigma2 Receptors. Bioorg. Med. Chem. 2008, 16, 755-61. 29. Geldenhuys, W. J.; Novotny, N.; Malan, S. F.; Van der Schyf, C. J., 3D-QSAR and Docking Studies of Pentacycloundecylamines at the Sigma-1 (Sigma1) Receptor. Bioorg Med Chem Lett 2013, 23, 1707-11. 30. Oberdorf, C.; Schmidt, T. J.; Wunsch, B., 5D-QSAR for Spirocyclic Sigma1 Receptor Ligands by Quasar Receptor Surface Modeling. Eur J Med Chem 2010, 45, 3116-24. 31. Huang, Y.; Hammond, P. S.; Whirrett, B. R.; Kuhner, R. J.; Wu, L.; Childers, S. R.; Mach, R. H., Synthesis and Quantitative Structure-Activity Relationships of N-(1-Benzylpiperidin-4-Yl)Phenylacetamides and Related Analogues as Potent and Selective Sigma1 Receptor Ligands. J Med Chem 1998, 41, 2361-70. 32. Oliveira, A. A.; Lipinski, C. F.; Pereira, E. B.; Honorio, K. M.; Oliveira, P. R.; Weber, K. C.; Romero, R. A. F.; de Sousa, A. G.; da Silva, A. B. F., New Consensus Multivariate Models Based on PLS and ANN Studies of Sigma-1 Receptor Antagonists. J. Mol. Model. 2017, 23, 302. 33. Avram, S.; Borcan, F.; Borcan, L. C.; Milac, A. L.; Mihailescu, D., QSAR Approaches Applied to Antidepressants Induced Neurogenesis--in Vivo and in Silico Applications. Mini Rev. Med. Chem. 2015, 16, 23040. 34. Laurini, E.; Marson, D.; Dal Col, V.; Fermeglia, M.; Mamolo, M. G.; Zampieri, D.; Vio, L.; Pricl, S., Another Brick in the Wall. Validation of the Sigma1 Receptor 3D Model by Computer-Assisted Design, Synthesis, and Activity of New Sigma1 Ligands. Mol. Pharm. 2012, 9, 3107-26. 35. Laurini, E.; Col, V. D.; Mamolo, M. G.; Zampieri, D.; Posocco, P.; Fermeglia, M.; Vio, L.; Pricl, S., Homology Model and Docking-Based Virtual Screening for Ligands of the Sigma1 Receptor. ACS Med. Chem. Lett. 2011, 2, 834-9. 36. Law, V.; Knox, C.; Djoumbou, Y.; Jewison, T.; Guo, A. C.; Liu, Y.; Maciejewski, A.; Arndt, D.; Wilson, M.; Neveu, V.; Tang, A.; Gabriel, G.; Ly, C.; Adamjee, S.; Dame, Z. T.; Han, B.; Zhou, Y.; Wishart, D. S., DrugBank 4.0: Shedding New Light on Drug Metabolism. Nucleic Acids Res. 2014, 42, D1091-D1097.



Page 30 of 33

37. Harder, E.; Damm, W.; Maple, J.; Wu, C.; Reboul, M.; Xiang, J. Y.; Wang, L.; Lupyan, D.; Dahlgren, M. K.; Knight, J. L.; Kaus, J. W.; Cerutti, D. S.; Krilov, G.; Jorgensen, W. L.; Abel, R.; Friesner, R. A., OPLS3: A Force Field Providing Broad Coverage of Drug-Like Small Molecules and Proteins. J. Chem. Theory Comput. 2016, 12, 281296. 38. Qiu, D.; Shenkin, P. S.; Hollinger, F. P.; Still, W. C., The GB/SA Continuum Model for Solvation. A Fast Analytical Method for the Calculation of Approximate Born Radii. J. Phys. Chem. A 1997, 101, 3005-3014. 39. Friesner, R. A.; Murphy, R. B.; Repasky, M. P.; Frye, L. L.; Greenwood, J. R.; Halgren, T. A.; Sanschagrin, P. C.; Mainz, D. T., Extra Precision Glide: Docking and Scoring Incorporating a Model of Hydrophobic Enclosure for Protein-Ligand Complexes. J Med Chem 2006, 49, 6177-96. 40. Dixon, S. L.; Smondyrev, A. M.; Knoll, E. H.; Rao, S. N.; Shaw, D. E.; Friesner, R. A., Phase: A New Engine for Pharmacophore Perception, 3D QSAR Model Development, and 3d Database Screening: 1. Methodology and Preliminary Results. J. Comput. Aided Mol. Des. 2006, 20, 647-671. 41. Lipinski, C. A.; Lombardo, F.; Dominy, B. W.; Feeney, P. J., Experimental and Computational Approaches to Estimate Solubility and Permeability in Drug Discovery and Development Settings. Adv Drug Deliv Rev 2001, 46, 3-26. 42. Hayashi, T.; Su, T. P., Sigma-1 Receptor Ligands: Potential in the Treatment of Neuropsychiatric Disorders. CNS Drugs 2004, 18, 269-84. 43. Matsumoto, R. R.; McCracken, K. A.; Friedman, M. J.; Pouw, B.; De Costa, B. R.; Bowen, W. D., Conformationally Restricted Analogs of BD1008 and an Antisense Oligodeoxynucleotide Targeting Sigma1 Receptors Produce Anti-Cocaine Effects in Mice. Eur. J. Pharmacol. 2001, 419, 163-74. 44. Diaz, J. L.; Cuberes, R.; Berrocal, J.; Contijoch, M.; Christmann, U.; Fernandez, A.; Port, A.; Holenz, J.; Buschmann, H.; Laggner, C.; Serafini, M. T.; Burgueno, J.; Zamanillo, D.; Merlos, M.; Vela, J. M.; Almansa, C., Synthesis and Biological Evaluation of the 1-Arylpyrazole Class of Sigma(1) Receptor Antagonists: Identification of 4-{2-[5-Methyl-1-(Naphthalen-2-Yl)-1h-Pyrazol-3-Yloxy]Ethyl}Morpholine (S1RA, E-52862). J Med Chem 2012, 55, 8211-24. 45. Nakazato, A.; Ohta, K.; Sekiguchi, Y.; Okuyama, S.; Chaki, S.; Kawashima, Y.; Hatayama, K., Design, Synthesis, Structure−Activity Relationships, and Biological Characterization of Novel Arylalkoxyphenylalkylamine σ Ligands as Potential Antipsychotic Drugs. J Med Chem 1999, 42, 1076-1087. 46. Itzhak, Y.; Stein, I.; Zhang, S. H.; Kassim, C. O.; Cristante, D., Binding of Sigma-Ligands to C57BL/6 Mouse Brain Membranes: Effects of Monoamine Oxidase Inhibitors and Subcellular Distribution Studies Suggest the Existence of Sigma-Receptor Subtypes. J. Pharmacol. Exp. Ther. 1991, 257, 141. 47. Gaulton, A.; Bellis, L. J.; Bento, A. P.; Chambers, J.; Davies, M.; Hersey, A.; Light, Y.; McGlinchey, S.; Michalovich, D.; Al-Lazikani, B.; Overington, J. P., Chembl: A Large-Scale Bioactivity Database for Drug Discovery. Nucleic Acids Res. 2012, 40, D1100-7. 48. Ganter, B.; Tugendreich, S.; Pearson, C. I.; Ayanoglu, E.; Baumhueter, S.; Bostian, K. A.; Brady, L.; Browne, L. J.; Calvin, J. T.; Day, G. J.; Breckenridge, N.; Dunlea, S.; Eynon, B. P.; Furness, L. M.; Ferng, J.; Fielden, M. R.; Fujimoto, S. Y.; Gong, L.; Hu, C.; Idury, R.; Judo, M. S.; Kolaja, K. L.; Lee, M. D.; McSorley, C.; Minor, J. M.; Nair, R. V.; Natsoulis, G.; Nguyen, P.; Nicholson, S. M.; Pham, H.; Roter, A. H.; Sun, D.; Tan, S.; Thode, S.; Tolley, A. M.; Vladimirova, A.; Yang, J.; Zhou, Z.; Jarnagin, K., Development of a Large-Scale Chemogenomics Database to Improve Drug Candidate Selection and to Understand Mechanisms of Chemical Toxicity and Action. J. Biotechnol. 2005, 119, 219-44. 49. Ganapathy, M. E.; Prasad, P. D.; Huang, W.; Seth, P.; Leibach, F. H.; Ganapathy, V., Molecular and LigandBinding Characterization of the Sigma-Receptor in the Jurkat Human T Lymphocyte Cell Line. J. Pharmacol. Exp. Ther. 1999, 289, 251-60. 50. Jaworska, J. S.; Comber, M.; Auer, C.; Van Leeuwen, C. J., Summary of a Workshop on Regulatory Acceptance of (Q)SARs for Human Health and Environmental Endpoints. Environ. Health Perspect. 2003, 111, 1358-60.


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


51. Tropsha, A., Best Practices for QSAR Model Development, Validation, and Exploitation. Mol Inform 2010, 29, 476-88. 52. Cherkasov, A.; Muratov, E. N.; Fourches, D.; Varnek, A.; Baskin, II; Cronin, M.; Dearden, J.; Gramatica, P.; Martin, Y. C.; Todeschini, R.; Consonni, V.; Kuz'min, V. E.; Cramer, R.; Benigni, R.; Yang, C.; Rathman, J.; Terfloth, L.; Gasteiger, J.; Richard, A.; Tropsha, A., QSAR Modeling: Where Have You Been? Where Are You Going To? J Med Chem 2014, 57, 4977-5010. 53. Peng, Y.; Keenan, S. M.; Zhang, Q.; Kholodovych, V.; Welsh, W. J., 3D-QSAR Comparative Molecular Field Analysis on Opioid Receptor Antagonists: Pooling Data from Different Studies. J Med Chem 2005, 48, 1620-9. 54. Peng, Y.; Keenan, S. M.; Zhang, Q.; Welsh, W. J., 3D-QSAR Comparative Molecular Field Analysis on Delta Opioid Receptor Agonist Snc80 and Its Analogs. J. Mol. Graph. Model. 2005, 24, 25-33. 55. Cao, X.; Chen, Y.; Zhang, Y.; Lan, Y.; Zhang, J.; Xu, X.; Qiu, Y.; Zhao, S.; Liu, X.; Liu, B. F.; Zhang, G., Synthesis and Biological Evaluation of Novel Sigma1 Receptor Ligands for Treating Neuropathic Pain: 6Hydroxypyridazinones. J Med Chem 2016, 59, 2942-61. 56. Sun, H.; Shi, M.; Zhang, W.; Zheng, Y. M.; Xu, Y. Z.; Shi, J. J.; Liu, T.; Gunosewoyo, H.; Pang, T.; Gao, Z. B.; Yang, F.; Tang, J.; Yu, L. F., Development of Novel Alkoxyisoxazoles as Sigma-1 Receptor Antagonists with Antinociceptive Efficacy. J Med Chem 2016, 59, 6329-43. 57. Lan, Y.; Chen, Y.; Xu, X.; Qiu, Y.; Liu, S.; Liu, X.; Liu, B. F.; Zhang, G., Synthesis and Biological Evaluation of a Novel Sigma-1 Receptor Antagonist Based on 3,4-Dihydro-2(1H)-Quinolinone Scaffold as a Potential Analgesic. Eur J Med Chem 2014, 79, 216-30. 58. Lan, Y.; Songyang, Y.; Zhang, L.; Peng, Y.; Song, J., Synthesis and Biological Evaluation of Novel 6,7Dihydro-5H-Cyclopenta[d]Pyrimidine and 5,6,7,8-Tetrahydroquinazoline Derivatives as Sigma-1 (Sigma1) Receptor Antagonists for the Treatment of Pain. Bioorg Med Chem Lett 2016, 26, 2051-6. 59. Lan, Y.; Chen, Y.; Cao, X.; Zhang, J.; Wang, J.; Xu, X.; Qiu, Y.; Zhang, T.; Liu, X.; Liu, B. F.; Zhang, G., Synthesis and Biological Evaluation of Novel Sigma-1 Receptor Antagonists Based on Pyrimidine Scaffold as Agents for Treating Neuropathic Pain. J Med Chem 2014, 57, 10404-23. 60. Chemi, G.; Gemma, S.; Campiani, G.; Brogi, S.; Butini, S.; Brindisi, M., Computational Tool for Fast in Silico Evaluation of Herg K(+) Channel Affinity. Front Chem 2017, 5, 7. 61. Mahajan, P.; Chashoo, G.; Gupta, M.; Kumar, A.; Singh, P. P.; Nargotra, A., Fusion of Structure and Ligand Based Methods for Identification of Novel CDK2 Inhibitors. J. Chem. Inf. Model. 2017, 57, 1957-1969. 62. Therese, P. J.; Manvar, D.; Kondepudi, S.; Battu, M. B.; Sriram, D.; Basu, A.; Yogeeswari, P.; Kaushik-Basu, N., Multiple E-Pharmacophore Modeling, 3D-QSAR, and High-Throughput Virtual Screening of Hepatitis C Virus NS5B Polymerase Inhibitors. J. Chem. Inf. Model. 2014, 54, 539-52. 63. Palakurti, R.; Sriram, D.; Yogeeswari, P.; Vadrevu, R., Multiple E-Pharmacophore Modeling Combined with High-Throughput Virtual Screening and Docking to Identify Potential Inhibitors of Beta-Secretase(Bace1). Mol Inform 2013, 32, 385-98. 64. Gilligan, P. J.; Cain, G. A.; Christos, T. E.; Cook, L.; Drummond, S.; Johnson, A. L.; Kergaye, A. A.; McElroy, J. F.; Rohrbach, K. W.; Schmidt, W. K.; et al., Novel Piperidine Sigma Receptor Ligands as Potential Antipsychotic Drugs. J Med Chem 1992, 35, 4344-61. 65. Zampieri, D.; Mamolo, M. G.; Laurini, E.; Florio, C.; Zanette, C.; Fermeglia, M.; Posocco, P.; Paneni, M. S.; Pricl, S.; Vio, L., Synthesis, Biological Evaluation, and Three-Dimensional in Silico Pharmacophore Model for Sigma(1) Receptor Ligands Based on a Series of Substituted Benzo[d]Oxazol-2(3H)-One Derivatives. J Med Chem 2009, 52, 5380-93. 66. Glennon, R. A.; Ablordeppey, S. Y.; Ismaiel, A. M.; el-Ashmawy, M. B.; Fischer, J. B.; Howie, K. B., Structural Features Important for Sigma 1 Receptor Binding. J Med Chem 1994, 37, 1214-9. 67. Laggner, C.; Schieferer, C.; Fiechtner, B.; Poles, G.; Hoffmann, R. D.; Glossmann, H.; Langer, T.; Moebius, F. F., Discovery of High-Affinity Ligands of Sigma1 Receptor, ERG2, and Emopamil Binding Protein by Pharmacophore Modeling and Virtual Screening. J Med Chem 2005, 48, 4754-64.



Page 32 of 33

68. Glennon, R. A., Pharmacophore Identification for Sigma-1 (Sigma1) Receptor Binding: Application of the "Deconstruction-Reconstruction-Elaboration" Approach. Mini Rev. Med. Chem. 2005, 5, 927-40. 69. Golbraikh, A.; Tropsha, A., Beware of Q2! J. Mol. Graph. Model. 2002, 20, 269-76. 70. Tropsha, A., Best Practices for QSAR Model Development, Validation, and Exploitation. Molecular Informatics 2010, 29, 476-488. 71. Guitart, X.; Codony, X.; Ballarín, M.; Dordal, A.; Farré, A. J., E-5842: A New Potent and Preferential Sigma Ligand. Preclinical Pharmacological Profile. CNS Drug Rev 1998, 4, 201-224. 72. Tam, S. W.; Steinfels, G. F.; Gilligan, P. J.; Schmidt, W. K.; Cook, L., DuP 734 [1-(Cyclopropylmethyl)-4(2'(4''-Fluorophenyl)-2';-Oxoethyl)- Piperidine Hbr], a Sigma and 5-Hydroxytryptamine 2 Receptor Antagonist: Receptor-Binding, Electrophysiological and Neuropharmacological Profiles. J. Pharmacol. Exp. Ther. 1992, 263, 1167. 73. Carr, K. D.; Hiller, J. M.; Simon, E. J., Diphenhydramine Potentiates Narcotic but Not Endogenous Opioid Analgesia. Neuropeptides 1985, 5, 411-4. 74. Chan, Y. T.; Hays, E. E., A Resin Complex for Prolonged Antitussive Effects. Am. J. Med. Sci. 1957, 234, 207-212.


Page 33 of 33

For Table of Contents use only

Pharmacophore-based 3D-QSAR

Virtual Screening

10 Predicted pKi

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


9 8 7 6 5 5

7 9 Experimental pKi

Diphenhydramine Ki=58 nM (S1R) Phenyltoloxamine Ki=160 nM (S1R)


Comprehensive 3D-QSAR Model Predicts Binding Affinity of

Recommend Documents