New role of phenothiazine derivatives as peripherally acting CB1 receptor antagonizing anti-obesity agents

Developing peripherally active cannabinoid 1 (CB1) receptor antagonists is a novel therapeutic approach for the management of obesity. An unusual phenothiazine scaffold containing CB1R antagonizing hit was identified by adopting virtual screening work flow. The hit so identified was further modified by introducing polar functional groups into it to enhance the polar surface area and decrease the hydrophobicity of the resulting molecules. CB1 receptor antagonistic activity for the designed compounds was computed by the previously established pharmacophore and three dimensional quantitative structure–activity relationship models. Docking studies of these designed compounds confirmed the existence of favourable interactions within the active site of the CB1 receptor. The designed compounds were synthesized and evaluated for their CB1 receptor antagonistic activity. Parallel artificial membrane permeability assay was performed to evaluate their potential to permeate into the central nervous system wherein it was observed that the compounds did not possess the propensity to cross the blood brain barrier and would be devoid of central nervous system side effects. In pharmacological evaluation, the synthesized compounds (23, 25, 27 and 34) showed significant decrease in food intake suggesting their potential application in the management of obesity through CB1 receptor antagonist activity.

Some peripherally acting selective CB1 antagonists such as AM-6545 (2) 49 synthesized at Center for Drug Discovery, Northeastern University, in the Alexander Makriyannis lab showed limited brain penetration with less neuropsychiatric side effects. 7TM Pharma reported TM38837 (3) 50 having low brain-to-plasma ratio (1:33) and potent CB1R antagonist activity (EC 50 = 8.5 nM). Another compound JD-5037 (4) 42 from Jenrin Discovery has successfully completed preclinical evaluation and the company is planning to file an investigational new drug (IND) application for it in 2017 as a peripherally acting CB1 receptor antagonist (Fig. 1).
Our group has been actively engaged in the design and development of novel peripherally acting CB1 receptor antagonists 3,13,51,52 . In the present work, we are reporting a novel series of compounds possessing phenothiazine scaffold as potential peripherally acting CB1 receptor antagonists. In the literature, compounds bearing phenothiazine scaffold have been reported to show different activities such as antipsychotic, anticonvulsant, antihistaminic, anticholinergic, antipruritic, antiemetic, antifungal, antibacterial, antiviral, anticancer, antimalarial, analgesic, anti-inflammatory, immunosuppressive, antifilarial, trypanocidal, and multidrug resistance reversal properties [53][54][55] . Probably this is the first report in which compounds with phenothiazine scaffold have been reported as peripherally acting CB1 receptor antagonists. The selection of phenothiazine scaffold was based on our previous communication 52 wherein we reported 14 hits obtained through molecular modeling studies. To our surprise, out of these 14 hits, one was having phenothiazine scaffold, which was quite unexpected as there were no reports on phenothiazine scaffold-containing compounds as CB1 receptor antagonists. The literature is flooded with monotonous reports 3,13,34-46 on vicinal diaryl heterocyclic systems exhibiting CB1 receptor antagonizing properties, and any other new chemical framework showing such an activity has the potential to give the drug discovery program on CB1 receptor antagonists, a new lease of life. Thus, it was planned to take up the synthesis of this novel hit 2-(2-(2-(trifluoromethyl)-10H-phenothiazin-10-yl)-2-oxoethylthio)-4-aminopyrimid ine-5-carbonitrile (V11) (Fig. 2) and evaluate it biologically. The hit (V11) (compound 5) so obtained was theoretically optimized by substituting different hydrophobic and hydrophilic functional groups, in light of the inputs obtained from the previously developed pharmacophore model and atom-based three dimensional quantitative structure-activity relationship (3D-QSAR) model. Thus, a series of phenothiazine derivatives were designed and synthesized which offered encouraging results in molecular modeling studies, in parallel artificial membrane permeability assay (PAMPA) and preliminary in vivo animal studies.

Results and Discussion
Identification of hit (V11) as anti-obesity agent. In the current era of drug discovery process, computational methods play a vital role in the identification of novel hits, which could be optimized further into clinically useful therapeutic agents. In our search for newer scaffolds as CB1 receptor antagonist, we successfully reported 52 new hits by applying 'virtual screening technique' on Asinex database (containing 435,214 compounds) using different filters and tools such as pharmacophore map, 3D-QSAR model, Lipinski's rule of five, CNS scoring and receptor-ligand interaction studies. Out of the initial 435,214 compounds, 14 compounds were finally identified as novel hits. Among these 14 hits, seven hits (V1, V4, V7, V8, V11, V12 and V14) were found to possess new scaffolds, which were never reported as CB1 receptor antagonists. Except for the hit (V11), the remaining six hits were having quite similar-type of scaffolds containing diaryl rings. The hit (V11) contained phenothiazine scaffold which has never been reported in the literature exhibiting CB1 receptor antagonist activity.
The hit (V11) was evaluated qualitatively using our previously developed pharmacophore model (AHRR) for peripherally acting CB1R antagonists. In the four featured pharmacophore (AHRR) model, the two phenyl rings of the phenothiazine scaffold of the hit (V11) occupied the aromatic ring (R12 and R13) features. The hydrophobic (H9) feature was occupied by the trifluoromethyl group attached to 2 nd position of the phenothiazine ring. The hydrogen bond acceptor (A5) feature was occupied by one of the nitrogen atoms of the pyrimidine ring indicating that the hit (V11) fitted well into the pharmacophore model as shown in Fig. 2b. The in silico activity of the obtained hit (V11) was computed by using the previously developed atom-based 3D-QSAR model 52 offering a pK i value of 7.55. The threshold value for the active compounds was set to be pKi ≥ 7.5 in the model indicating that the hit (V11) could be an important true positive. Orientation of the hit (V11) in the active site of CB1 receptor was also studied and was found to be quite similar to rimonabant (Fig. 2c). The trifluoromethyl group attached to the 2 nd position of the phenothiazine ring of the hit (V11) was oriented towards Val196, Phe200, Trp356 and Leu360 residues similar to the 4-chlorophenyl group of rimonabant. Ring A of the phenothiazine scaffold of the hit (V11) was oriented towards Trp279 and Met363 residues similar to the rimonabant's 2,4-dichlorophenyl ring. The pyrimidine ring of the hit (V11) was oriented towards Phe177, Phe189 and Trp255 residues similar to the piperidyl ring of rimonabant. The hit (V11) made a hydrogen bond attachment between the nitrogen of pyrimidine ring and Lys192, which was considered to be essential for exhibiting CB1 receptor antagonistic activity. Thus, the docking studies indicated that the hit (V11) was having an orientation similar to rimonabant in the active site of the CB1 receptor as shown in Fig. 2c. The results of the pharmacophore mapping, 3D-QSAR and docking studies clearly indicated that the hit (V11) was an appropriate case for further structural optimization. The result of biological evaluation of the hit (V11) (compound 5) as an orally active CB1 receptor antagonist supported the theoretical studies. So, it was planned to go ahead with the synthesis and optimization of a novel series of phenothiazines as peripherally acting CB1 receptor antagonist for the treatment of obesity.
Optimization of the hit V11 (5). CB1 receptor antagonists are believed to be potent anti-obesity agents.
But, their clinical application is limited because of their associated central side effects. To overcome their CNS side effects, the clinical candidates must have restricted entry into the brain. Therefore, the current drug designing approach was aimed at developing peripherally active compounds only, as successful anti-obesity agents.
In order to enhance the potency and PSA, and decrease the hydrophobicity, in silico modifications were carried out in the structure of the hit (V11) molecules, and the resulting compounds were evaluated initially using molecular modeling techniques. In accordance to the pharmacophore (AHRR) model, two phenyl rings of the phenothiazine scaffold of the hit (V11) were perfectly aligned on the pharmacophoric ring features (R12 and R13), hence it was decided not to temper with the phenyl rings of the phenothiazine scaffold in the new molecules to be designed. Modifications were envisaged on the other two pharmacophoric sites in the molecule i.e. on the sites having the hydrophobic feature and the hydrogen bond acceptor feature. The hydrophobic (H) feature of the pharmacophore was occupied by the trifluoromethyl group of the hit (V11). So, different hydrophobic substituents such as trifluoromethyl, chloro and methoxy were planned to be attached, at this position and the activity of the resulting compounds was predicted. Keeping hydrogen at this position decreased the activity as was found later on in case of compound (24) because the molecule complied only with three pharmacophoric features (ARR). In the pharmacophoric region having the hydrogen bond acceptor (A) group, it was decided to attach some polar functional groups so that these groups could occupy the hydrogen bond acceptor region that would show mandatory interaction with the Lys192 residue in the active site, a key amino acid residue for CB1 receptor binding. Thus, considering the synthetic feasibility, in silico assessment of compounds having polar functional groups such as 4,5-dihydrothiazole, 1-phenyl-1H-tetrazole, 1,3,4-thiadiazol-2-amine, 4-(1,3,4-oxadiazol-2-yl) pyridine, 4-amino-5-(4-pyridyl)-4H-1,2,4-triazole and ethyl 4-aminopyrimidine-5-carboxylate, was done. The 4,5-dihydrothiazole substituted compound showed low PSA and G-score values of 33.15 Å 2 and -8.58 respectively in the in silico studies. In 1-phenyl-1H-tetrazole substituted compound, the PSA increased to 69.66 Å 2 but the G-score got affected adversely (−7.44). The 1,3,4-thiadiazol-2-amino substituted compound offered comparatively better values of PSA and G-score i.e. 74.10 Å 2 and −8.85 respectively. 4-(1,3,4-Oxadiazol-2-yl)pyridine substituted compounds gave improved PSA value of 76.81 Å 2 and G-score of −10.79. Substitution with 4-amino-5-(4-pyridyl)-4H-1,2,4-triazole group increased the PSA value to 96.27 Å 2 and improved the G-score further to −11.74. The ethyl 4-aminopyrimidine-5-carboxylate substituted compound showed the highest PSA value of 106.26 Å 2 with a good G-score of −10.37. Higher values of PSA and more negative G-scores in the docking experiments suggested that the substituted pyrimidine, oxadiazole and triazole groups could prove to be favourable for peripheral CB1R antagonist activity. Furthermore, good fitness scores and improved predictive activities were obtained for these compounds in the pharmacophore modeling and atom-based 3D-QSAR studies respectively; therefore they were selected for chemical synthesis.
Efforts were also made to vary the length of the carbon chain spacer between the phenothiazine scaffold and the H-bond acceptor group like 4-amino-2-mercaptopyrimidine-5-carbonitrile. The two carbon chain-bearing compounds yielded poor results (fitness score = 1.09, pKi = 6.90, G score = −9.17) in comparison to the compounds having three carbon spacer (fitness score = 1.15, pKi = 8.20, G score = −10.17). So, it was decided to synthesize some compounds having three carbon-chain spacer also. The fitness score, predicted activity (pK i ) and G-scores of the designed compounds are shown in Table 1.
So, the identified hit (V11) was modified by incorporating different hydrophobic and polar functional groups in its structure. The compounds planned for the synthesis contained groups such as trifluoromethyl, chloro, methoxy and hydrogen, at 2 nd position of the phenothiazine ring and polar functional groups such as substituted pyrimidine, oxadiazole and triazole attached to the amide chain. Some compounds having a spacer of three carbon chain in between the phenothiazine scaffold and the polar pyrimidine ring were also planned to be synthesized. All the designed compounds were evaluated in silico using the pharmacophore model, 3D-QSAR model and docking studies, prior to their synthesis, as shown in Table 1.

G-score (Docking)
The physicochemical and pharmacokinetic parameters of the designed compounds were also calculated. Physicochemical properties such as molecular weight (391.46-506.56 Da), number of hydrogen bond donors (0-2), number of hydrogen bond acceptors (6-9), hydrophobicity (2.77-6.31), polar surface area (75.22-114.55) and number of rotatable bonds (3)(4)(5)(6)(7)(8)(9) were all in the acceptable range. All of the designed compounds were theoretically predicted to show good absorption (71.62-100%) in humans. BBB permeability, an important parameter for peripherally acting CB1 antagonists, was also predicted for the designed compounds. The threshold value for the compounds which could readily cross the BBB was logBB > 0.3 whereas logBB < −1 indicated poor penetration into the brain 56 . The computed values of BBB permeability of the designed compounds were in the range −0.52 to −1.72, whereas the BBB permeability value for rimonabant was calculated to be 0.23. Based upon this data, it was expected that the designed compounds would not cross the BBB. It was also predicted that all the compounds would be metabolized by CYP3A4 enzyme system. The total clearance of the synthesized compounds, ranging from 0.009-0.351 ml/min/kg, indicated good physicochemical and pharmacokinetic properties ( Table 2) suggesting their drug-like behavior.
Biological. In vitro permeability assay of the test compounds. As discussed above, BBB permeation of the available CB1 receptor antagonists limits their therapeutic utility. A successful CB1 receptor antagonist to be used as anti-obesity agent must be devoid of BBB permeability. Compounds which showed predicted activity (pKi) of ≥7.5 were selected for evaluation for their tendency to penetrate the BBB. In this study, the ability of the test compounds to cross the BBB was determined by the well established in vitro model PAMPA-BBB assay. This simple and rapid model has the advantage to predict passive BBB permeation with high accuracy 57,58 . The in vitro permeability (P e ) of the test compounds through the lipid extract of porcine brain was determined in PBS/ethanol (70:30). All the test compounds showed permeability (P e ) values lesser than the threshold value of 3.8 × 10 -6 cm s −1 , suggesting 58 that they would have nil/low propensity to cross the BBB by passive diffusion (Table 3).    (Fig. 5A).
In another set of experiment, the selected test compounds (23, 25, 27 and 34) were further evaluated to assess their in vivo CB1 receptor antagonistic activity in presence of WIN-55212-2, a potent cannabinoid receptor agonist. Here also, the quantity of feed consumed by the control group animals was taken as 100%. The animals treated with the CB1R agonist WIN-55212-2, consumed 154% feed in comparison to the control group animals. Ability of the test compounds to act as CB1 receptor antagonists was assessed by their capability to attenuate the hyperphagic response shown by the animals treated with WIN-55212-2. WIN-55212-2 significantly increased (54%) the amount of food intake in treated animals as compared to the control group (p < 0.05). However, the hyperphagic effect of WIN-55212-2 was significantly attenuated by the test compounds (23, 25, 27 and 34) (p < 0.001). These compounds (23, 25, 27 and 34) showed significant decrease (36%, 36%, 26% and 28%, respectively) in food intake as compared to the control. The results supported CB1 receptor antagonistic potential of the test compounds (23, 25, 27 and 34) (Fig. 5B).    (32) showed a lowered fitness score of 1.14 as it did not contain a hydrophobic group at 2 nd position of the phenothiazine ring when mapped in the pharmacophore model. Absence of the hydrophobic group flipped the molecule in such a way that one of the aromatic features was occupied by the phenyl ring of the phenothiazine scaffold and the other one by the oxadiazole ring. Hydrogen bond acceptor feature was oriented near the nitrogen of the pyridine ring in the attached side chain as shown in Fig. 6b. So, the orientation of compound (32) got a lowered fitness score in the pharmacophore mapping. It became clear from this study that it was necessary for the phenothiazine derivatives to match all the pharmacophoric features for good biological activity.
Docking studies. In docking studies, it was observed that formation of a hydrogen bond between Lys192 residue of the active site and the docked molecule was a characteristic feature for binding to the CB1 receptor. The virtual screening hit V11 (5) molecule having a G-score of −10.58 was observed to have the required binding in the active site of CB1 receptor with the Lys192 residue. Compounds (25 and 27) exhibiting G-scores of −10.37 and −10.33 respectively, formed two hydrogen bonds, one between one of the nitrogen atoms of the pyrimidine ring and the Lys192 residue, and another between the hydrogen of the amine attached to pyrimidine ring and the Ser265 residue, which might be responsible for their good biological activity. The orientation of compound (25) in the active site of CB1 receptor is shown in Fig. 7a. Similarly, compounds (23 and 34) also exhibited good G-scores (−10.04 and −9.80 respectively) and were suitably well oriented in the active site of CB1 receptor, which supported their good biological activity. The lowest G-score (−8.70) was obtained for compound (47). Although  it also formed a hydrogen bond with Lys192 residue as shown in Fig. 7b but still showed a comparatively low G-score. This might be due to the presence of three carbon chain spacer which increased the distance between the phenothiazine scaffold (containing two aromatic rings and one hydrophobic features) and the substituted pyrimidine ring (containing a hydrogen bond acceptor feature).
Structure activity relationship (SAR). Different phenothiazine derivatives were synthesized so that a preliminary structure activity relationship could be developed. For developing a robust SAR, a lot more number of compounds having different types of groups would be needed. In the acute hypophagic study, the food intake got decreased (62%) when the animals were treated with compound (5), the virtual screening hit (V11). Replacement of trifluoromethyl group of compound (5) with methoxy group yielded compound (23) which caused significant decrease in the food intake to 19% of the control. Replacement of 4-amino-2-mercaptopyrimidine-5-carbonitrile group of compound (5) by 4-aminopyrimidine-5-carboxylate resulted in compound (25) causing decrease in activity (food intake 37%). The trifluoromethyl group of compound (25) when replaced by methoxy group in compound (27) again caused a substantial improvement in the activity (food intake 19%). The trifluoromethyl and 4-amino-2-mercaptopyrimidine-5-carbonitrile groups of compound (5) when replaced by chloro and 4-amino-5-(4-pyridyl)-4H-1,2,4-triazole groups resulted in compound (34) having a lower activity (decreasing the food intake to 43%). Replacement of chloro group of 34 by trifluoromethyl group in compound (33) did not impact the food intake. Similarly, presence of hydrogen and 4-(1,3,4-oxadiazol-2-yl)pyridine groups in compound (32) also showed no effect on food intake. Removal of the amide linkage of compound (5) by three carbon methylene chain resulted in compound (42), again having no effect on the food intake. In another experiment carried out using WIN-55212-2 as CB1R agonist, compounds (23, 25, 27 and 34) showed significant decrease in food intake (36%, 36%, 26% and 28% respectively) which indicated their potential in vivo CB1 receptor antagonism. In nutshell, the SAR revealed that the hydrophobic substituents such as trifluoromethyl, chloro and methoxy at 2 nd position of phenothiazine ring were beneficial for the activity, especially the methoxy group in compounds (23 and 27). Whereas, absence of hydrophobic feature at 2 nd position of the phenothiazine scaffold resulted in poor activity for the resulting compounds as observed in compound (32), underlining the importance of a hydrophobic feature at this position. Amide linkage in the spacer was observed to be favourable for the activity. Replacement of the amide group spacer by a three carbon methylene chain spacer was observed to offer comparatively less active compounds (42-44, 46 and 47) than in comparison to the compounds (23, 25, 27 and 34) having an amide linkage. Different polar groups, such as 4-amino-2-mercaptopyrimidine-5-carbonitrile in compound (23), 4-aminopyrimidine-5-carboxylate in compounds (25 and 27) and 4-amino-5-(4-pyridyl)-4H-1,2,4-triazole group in compound (34), were found useful to increase the PSA of the resulting compounds and also acted as hydrogen bond acceptor features which formed hydrogen bond with the Lys192 residue providing favourable CB1R binding affinity to the ligands.

Conclusion
With the aim to diversify the chemical domain of the existing peripherally acting CB1 antagonists, a novel hit possessing phenothiazine scaffold was identified. The hit molecule was structurally optimized to afford a novel series of phenothiazine derivatives having higher PSA and lower hydrophobicity as compared to rimonabant. Compounds of the designed series were synthesized. The synthesized compounds are hypothesized not to cross the blood brain barrier as assessed by PAMPA assay, indicating their peripherally restricted presence that could make them devoid of unwanted CNS side effects. In the preliminary screening, it was found that the hit V11 (5) possessed weak short acting hypophagic activity. Further modifications in the hit V11 (5) by using hydrophobic groups such as chloro and methoxy groups at 2 nd position of the phenothiazine ring were found beneficial for the activity in computational studies as well as whole animal assay. Substituted pyrimidine and triazole rings acted as hydrogen bond acceptors forming hydrogen bond with Lys192 residue, which was favourable for the activity. In the in vivo studies, the structurally modified compounds (23, 25, 27 and 34) showed a significant decrease in food intake in comparison to the control group. Thus, the phenothiazine moiety has been identified as a promising scaffold in providing peripherally acting CB1 receptor antagonists for the management of obesity with nil/minimum CNS side effects. To the best of our knowledge this is the first report wherein the phenothiazine scaffold has been shown to exhibit peripherally acting CB1 receptor antagonistic activity which could be exploited therapeutically for the management of obesity without affecting the psyche of the subject under treatment.

Experimental
Computational. All the computational studies reported in the present work were carried out using Maestro (version 9.0 and 9.4) software of Schrödinger, New York 59,60 . The previously built pharmacophore and atom-based 3D-QSAR models were used in this study 52 for mapping of the pharmacophoric features of the designed compounds, and prediction of the biological activity. For molecular modeling purposes, chemical structures of all the compounds were sketched, cleaned up and minimized by using OPLS 2005 force field in the standard tool 'Ligprep' version 2.3 61 . Molecular docking studies for the designed compounds were performed by using standard tool 'Glide' version 5.5 62 . The co-crystallized structure for CB1 receptor (PDB code: 5TGZ 63 ) was downloaded from RCSB Protein Data Bank. Protein preparation was done by all hydrogen atom addition and water molecules deletion. The hydrogen bond assignment was optimized by exhaustive sampling method and the obtained protein structure was energy minimized in presence of OPLS_2005 force field using Impref module in the standard option 'protein preparation wizard' of Maestro 9.0. The obtained protein structure was used for receptor grid generation 64  Chemistry. All the solvents and reagents used in the synthesis were of analytical reagent grade commercially procured from Sigma-Aldrich, Spectrochem, S. d. fine chemicals or Avra chemicals. Purification of all the solvents and reagents were carried out by using general laboratory techniques before their use for synthesis of the desired compounds 66 . The melting points (mp) of the compounds were determined by using Veego make silicon oil bath-type melting point apparatus and are uncorrected. Thin layer chromatography (TLC) using silica gel pre-coated plates (60F 254 , Merck, 0.25 mm thickness) were employed to monitor the progress of the reactions. TLC plates were visualized with ultraviolet light (254 nm) or iodine vapors. The practical yields of the compounds reported here are un-optimized. The infrared spectra (IR) expressed in wave numbers in cm −1 were obtained using BRUKER ALPHA-T (Germany) FT-IR spectrophotometer where potassium bromide discs were used in the region of 4000-450 cm −1 . 1 H-NMR (400 MHz) spectra and 13 C-NMR spectra were determined on Bruker Advance-II spectrometer in CDCl 3 or DMSO-d 6 solvents; the chemical shift has been expressed in parts per million (δ ppm) and coupling constant (J) in Hz. The mass spectra were taken either on Thermo Fisher mass spectrometer having EI ion source or Shimadzu LCMS2020 with APCI & ESI probes for the compounds. Chromatographic separations were carried out on columns using silica gel (100-200) as the adsorbent.

4-Amino-2-mercaptopyrimidine-5-carbonitrile (9). Method A:
Freshly cut sodium (0.19 g, 8.26 mmol) was added into absolute ethanol (5 mL) in cold conditions. To this solution, thiourea (8) (0.62 g, 8.15 mmol) was added at room temperature. When thiourea got dissolved, ethoxymethylenemalononitrile (6) (1 g, 8.19 mmol) in ethanol (5 mL) was added using dropping funnel to the above stirred mixture over a period of 2 h and completion of the reaction was monitored by thin layer chromatography (TLC). The solvent was evaporated, water (10 mL) added to the residue and 2 M hydrochloric acid (HCl) was added to make the solution acidic. The precipitates so obtained were filtered to obtain a yellow colored solid of compounds (9)
In another set of experiments, in vivo CB1 receptor antagonistic activity of the test compounds was assessed. WIN-55212-2, a selective CB1 receptor agonist is known to increase the food intake. To assess the CB1 receptor antagonistic activity of the test compounds, 24 h fasted animals were given WIN-55212-2 (2 mg/kg, i.p.), 15 min prior to the administration of the test compounds and the experiment was proceeded as described above.