QSAR and Docking Studies on Capsazepine Derivatives for Immunomodulatory and Anti-Inflammatory Activity

Capsazepine, an antagonist of capsaicin, is discovered by the structure and activity relationship. In previous studies it has been found that capsazepine has potency for immunomodulation and anti-inflammatory activity and emerging as a favourable target in quest for efficacious and safe anti-inflammatory drug. Thus, a 2D quantitative structural activity relationship (QSAR) model against target tumor necrosis factor-α (TNF-α) was developed using multiple linear regression method (MLR) with good internal prediction (r2 = 0.8779) and external prediction (r2 pred = 0.5865) using Discovery Studio v3.5 (Accelrys, USA). The predicted activity was further validated by in vitro experiment. Capsazepine was tested in lipopolysaccharide (LPS) induced inflammation in peritoneal mouse macrophages. Anti-inflammatory profile of capsazepine was assessed by its potency to inhibit the production of inflammatory mediator TNF-α. The in vitro experiment indicated that capsazepine is an efficient anti-inflammatory agent. Since, the developed QSAR model showed significant correlations between chemical structure and anti-inflammatory activity, it was successfully applied in the screening of forty-four virtual derivatives of capsazepine, which finally afforded six potent derivatives, CPZ-29, CPZ-30, CPZ-33, CPZ-34, CPZ-35 and CPZ-36. To gain more insights into the molecular mechanism of action of capsazepine and its derivatives, molecular docking and in silico absorption, distribution, metabolism, excretion and toxicity (ADMET) studies were performed. The results of QSAR, molecular docking, in silico ADMET screening and in vitro experimental studies provide guideline and mechanistic scope for the identification of more potent anti-inflammatory & immunomodulatory drug.


Introduction
Capsicum species commonly known as chillies, relished as an important spice in the culinary art of the world and is known for its medicinal effect since the dawn of the human civilization. The medicinal property of 'hot pepper' has been attributed to the presence of capsaicin, a pungent principal ingredient produced as a secondary metabolite. Chemically known as 8-methyl-N-vanillyl-6-nonenamide. Capsaicin and its related compounds, collectively referred as 'capsaicinoids' or 'vanilloids', which bind specifically to transient receptor cation channel subfamily V (TRPV), that carry sensation of pain and responds naturally to noxious stimuli like high temperature and acidic pH [1]. Prolonged exposure causes nociceptor terminals to become insensitive to capsaicin, as well to other noxious stimuli [2]. Hyper stimulation of TRPV1 by capsaicin has an analgesic effect, since it leads to long-term desensitization of the sensory neurons. The clinical uses of TRPV1 agonist like capsaicin, are limited due to side effects of a burning sensation, irritation and neurotoxicity [3]. On the other hand, blocking of the pain-signalling pathway with a TRPV1 antagonist capsazepine represents a promising strategy for the development of novel analgesics with potentially fewer side effects [4]. Several non-neuronal effects of capsaicin have also been reported viz., induction of apoptosis in transformed cells [5], stimulation of prostaglandin formation leading to inhibition of gastric lesion [6], antibacterial activity [7], inhibition of cardiac excitability [8] and platelet aggregation [9]. Capsazepine is a known analog of capsaicin, discovered as a result of structure-activity relationship (SAR) studies [10]. Capsazepine induced similar action as capsaicin and resiniferatoxin (RTX) and exhibits even twofold more potent inhibition of expression of iNOS gene in LPSstimulated murine macrophages through inactivation of NF-kB [11,12]. NF-kB is a protein complex that control transcription of DNA and it is involved in cellular responses to stimuli such as stress, cytokines, free radicals, ultraviolet irradiations, oxidized low-density lipoprotein, and microbial antigens. NF-kB regulation of immune response and inflammation, cell lineage development, cell apoptosis, cell cycle progression and oncogenesis in response to stimuli have been shown to regulate the expression of several genes (bcl-2, bcl-xl), cellular inhibitor of apoptosis protein, tumor necrosis factor signalling pathway-related regulatory factor, cyclooxygenase-2 (COX-2), matrix metalloprotein peptide-9 (MMP-9) and inducible nitric oxide synthase (iNOS) and those for cell cycle regulatory components involved in tumorigenesis [13][14][15][16][17][18]. Therefore, in this work we have investigated the chemo preventive potential of capsazepine and its derivatives against proinflammatory mediator TNF-a through QSAR, in vitro activity evaluation and molecular docking studies, to understand the mechanism of action of vanilloids against inflammation and immunomodulation related to cancer. QSAR modelling also furnished the activity dependent structural descriptors and predicts the effective dose of other derivatives, thereby suggesting the possible toxicity range. Drugability of hit compounds was evaluated by using Lipinski's 'Rule of Five' and ADMET analysis through bioavailability filters.

Dataset preparation
A total of 146 known TNF-a inhibitors were collected from various published literatures based on its structural diversity and activity coverage. The activity data for all compounds were taken from different scientific groups in terms of inhibitory activity (IC 50 mM) [19][20][21][22][23][24]. 124 compounds out of 146, were selected as a training set based on following criteria to produce a good quantitative QSAR model: by covering a wide activity range of compounds and by including the most active, moderate and less active inhibitors (Table S1 in File S1). The biological activity for TNF-a inhibitors were ranging between 0.09 to 50 mM. The remaining 22 compounds were used as a test set to validate the generated model (Table S2 in File S1).

Energy minimization
The structural drawing and geometry cleaning of the training set compounds were performed through, ChemBioOffice suite Ultra v12.0 (2010) software (CambridgeSoft Corp., UK). The compounds then subjected to energy minimization by using Discovery Studio v3.5 software (Accelrys Inc., USA) by applying CHARMm forcefield applicable to most of the small molecules. It adds several properties to the compounds including: initial potential energy, RMS gradient, CHARMm energy and minimization criteria.

Chemical descriptors calculation
Molecular descriptors were calculated for each compounds using ''Calculate Molecular Properties'' module of the Discovery Studio v3.5 (Accelrys Inc., USA). These descriptors include 2D parameters (e.g., AlogP, molecular weight, number of aromatic ring, number of H-acceptors, number of H-donors, number of rings, number of rotatable bonds, molecular fraction polar surface area) and 3D (Dipole and Jurs descriptor).

Quantitative structure activity relationship (QSAR) model development
The set of energy optimised 146 compounds with calculated molecular properties were used for QSAR model development using create QSAR model module in Discovery Studio v3.5. Firstly all compounds were prepared for QSAR, and then the biological activities were specified as dependent property. Compounds were randomly divided into training (124 compounds) and test (22 compounds) set. This division was performed in such a manner that data coordinates of regression graph represent both   training and test set compounds and distributed within the whole descriptor space of the entire dataset. Each data point of the test set showed closer match with the training set compounds. The regression model equation was derived by using statistical multiple linear regression approach (Table S3 in File S1).

Model quality assessment and validation
The successful QSAR model must be robust enough to make accurate and reliable predictions of the non-investigated or query set compounds, therefore the obtained QSAR model from the training set should be subsequently validated. The conventional validation strategy for QSAR model analysis, based on multiple linear regression, include the calculation of cross validated squared correlation coefficient (r 2 ) for internal validation and the predictive squared correlation coefficient (r 2 pred ) for external validation. Here in this case the r 2 was 0.878, and r 2 pred was 0.5865, which ultimately prove the true predictability of model and the  summarized in Table 1 and structure of active capsazepine derivatives are showed in Figure 2.

Ethics Statement
Primary macrophage cells were isolated from the peritoneal cavities of the healthy female Swiss albino mice as per the

Primary cell culture and treatment
Primary cell culture was carried out as described previously [25]. In brief, the macrophage cells were collected from the peritoneal cavities of mice (8-week-old female Swiss albino mice) after an intra-peritoneal (i.p.) injection of 1.0 mL of 1% peptone (BD Biosciences, USA) 3 days before harvesting. Mice were euthanized by cervical dislocation under ether anesthesia and peritoneal macrophages were obtained by intra-peritoneal injection of Phosphate Buffer Saline (PBS), pH-7.4. Membrane debris was removed by filtering the cell suspensions through sterile gauze. The viability of the cells was determined by trypan blue exclusion and the viable macrophage cells at the concentration of 0.56106 live cells/mL were used for the experimentation. The cells were suspended in RPMI 1640 medium (Sigma-Aldrich, USA) containing 10% heat-inactivated fetal calf serum (Gibco, USA), 100 mg/mL of penicillin and 100 mg/mL of streptomycin and incubated in a culture plate (Nunc, Germany) at 37uC in 5% CO 2 in an incubator. Non-adherent cells were removed after 4 h by removing the culture media and the adherent cells were re- suspended in RPMI 1640 medium containing 10% heat-inactivated fetal calf serum. Cells were pretreated with 1, 2.5, 5 and 10 mg/mL of test compounds and standard anti-inflammatory drug, Dexamethasone (Sigma Aldrich, USA) at 5 mg/mL for 30 min. The cells were stimulated with lipopolysaccharide (LPS, 0.5 mg/mL). After incubation with LPS for 24 h, supernatants were collected and immediately frozen at 280uC. Harvested supernatants were tested for quantification of pro-inflammatory mediator TNF-a by ELISA method according to the manufacturer's instructions (BD Biosciences, USA).

Quantification of pro-inflammatory cytokines
Quantification of TNF-a at protein level in cell culture supernatant was carried out using Enzyme Immuno Assay (EIA) kits from BD Biosciences, USA following the manufacturer's protocol. Briefly, the ELISA plates (96 well) were coated (100 mL per well) with specific mouse TNF-a capture antibody respectively and incubate overnight at 4uC. The plate was blocked with 200 mL/well assay diluents. Cell Culture supernatant and standard (100 mL) were added into the appropriate coated wells and incubated for 2 h at room temperature (20225uC). After incubation, the plates were washed thoroughly 5 times with wash buffer. 100 mL of detecting solution (detection antibody and streptavidin HRP) was added in to each well. Seal plate and incubate for 1 h at RT and then the plates were washed thoroughly 5 times with wash buffer. 100 mL of tetramethyl benzidine (TMB) substrate solution to each well and incubate plate (without plate sealer) for 30 min at room temperature in the dark. Add 50 mL of stop solution (2N H 2 SO 4 ) to each well. The color density was measured at 450 and 570 nm using a microplate reader (Molecular Devices, USA). Subtract absorbance at 570 nm from absorbance 450 nm. The values of TNF-a were expressed as mg/mL and the IC 50 values was calculated from vector defined by percentage inhibition values obtained against concentration gradient ranging from 1-10 mg/mL Representative results are depicted in Figure 3 and Table 2.

Statistical analysis
Results were presented as the means 6SEM and analyzed using GraphPad Prism 4. The ANOVA followed by turkeys multiple comparison tests was used to assess the statistical significance of vehicle verses treatment groups. Results are presented as the means 6SEM. Differences with a p value ,0.05 were considered significant. IC 50 values were calculated from vector defined by percentage inhibition values obtained against a concentration gradient ranging from 1210 mg/mL.

Molecular docking
The docking study of selected target and ligands was done by using Autodock Vina v0.8 (Molecular Graphics Lab at The Scripps Research Institute, La Jolla, CA 92037, USA). The 3D crystallographic structure of anti-inflammatory protein target tumor necrosis factor-a (TNF-a) was retrieved through Brookhaven Protein DataBank (PDB) (http://www.pdb.org) (PDB ID: 2AZ5). The crystallographic protein structure of TNF-a complexes with known inhibitor was selected for docking procedure validation by re-docking approach and also to know the standard docking energy and binding site. The valency and hydrogen bonds of the ligands as well as target protein was subsequently satisfied.  An extended PDB format, termed as PDBQT file was used for coordinate files that includes atomic partial charges. The software automatically convert PDB file into PDBQT that was further used for docking [27]. Polar hydrogen atoms were added to the protein target to achieve the correct ionisation and tautomeric states of amino acid residues such as HIS, ASP, SER and GLU. For software standardization, native ligand of co-crystallised complex was first extracted and re-docked to its corresponding binding site using AutoDock Vina v0. 8

Screening through pharmacokinetic properties
Most of drugs in development failed during clinical trials due to poor pharmacokinetics parameters [28,29]. These properties such as absorption, distribution, metabolism, excretion and toxicity (ADMET) are important in order to determine the success of the compound for human therapeutic use. Some important chemical descriptors correlate well with ADMET properties such as polar surface area (PSA) as a primary determinant of fraction absorption, low molecular weight (MW) for oral absorption. The distribution of the compound in the human body depends on factors such as blood-brain barrier (Log BB), permeability such as apparent Caco-2 permeability, apparent MDCK cell permeability, Log Kp for skin permeability, volume of distribution and plasma protein binding (Log Khsa for Serum protein binding). It has been reported that excretion process that eliminates the compound from human body depends on the molecular weight and octanol-water partition coefficient (LogP). Similarly, rapid renal clearance is associated with small and hydrophilic compounds. The metabolism of most drugs that takes place in the liver is associated with large and hydrophobic compounds. Higher lipophilicity of compounds leads to increased metabolism and poor absorption, along with an increased probability of binding to undesired hydrophobic macromolecules, thereby increasing the potential for toxicity. In spite of some observed exceptions to Lipinski's rule, the property values of the vast majority (90%) of the orally active compounds are within their cut-off limits [30231]. In addition, the bioavailability of derivatives was assessed through topological polar surface area analysis. We calculated the polar surface area Table 4. Comparison of binding affinity of capsazepine and its active derivatives in terms of docking energy and binding site residues against anti-inflammatory receptor TNF-a (PDB: 2AZ5). (PSA) by using method based on the summation of tabulated surface contributions of polar fragments termed as topological PSA (TPSA). Generally, it has been seen that passively absorbed compounds with a PSA.140 Å 2 are thought to have low oral bioavailability. Calculations of other important ADME properties of capsazepine derivatives were performed through Discovery Studio v3.5, USA (2013). We also screened capsazepine and its derivatives through TOPKAT toxicity estimation using Discovery Studio v3.5. TOPKAT computes a probable value of toxicity for a submitted chemical structure from a quantitative structure-toxicity relationship (QSTR) equation. The product of a structure descriptors and its corresponding coefficient is the descriptors contribution to the probable toxicity.

Results and Discussion
In the present work, derivatives of capsazepine were evaluated for their anti-inflammatory activity through the developed QSAR model and docking studies. Structure activity relationship has been denoted by QSAR model showing significant internal prediction (r 2 = 0.8779) and external prediction (r 2 pred = 0.5865) (Figure 1). A total of 124 known inhibitors of TNF-a were used for QSAR modeling against 57 2D and 3D chemical descriptors. Eight descriptors were found to be significantly responsible for antiinflammatory activity (Table 1). A forward feed multiple linear regression QSAR model was developed. Both internal and external validations were performed for the developed model. A low residual value for each compound in the dataset defines the degree of correlation between observed and predicted values and the models predictive ability. Screening of derivatives through developed QSAR model indicated that derivatives CPZ-29, CPZ-30, CPZ-33, CPZ-34, CPZ-35 and CPZ-36 showed significant activity in compared to capsazepine's in vitro IC 50 against TNF-a ( Table 2). In 1989, it was found that capsazepine and some of its derivatives possessed extraordinary anti-inflammatory activity against COX-2 and inducible nitric oxide (iNO) [11,12]. In the present work, we report anti-inflammatory activity of 44 virtually designed capsazepine derivatives with lactone ring pharmacophore against pro-inflammatory target TNF-a. The activity of newly designed derivatives were predicted through the developed QSAR model and the derivatives CPZ-33 and CPZ-34 found to be better in activity as compared to capsazepine, whereas CPZ-30 was found close to capsazepine. All six derivatives and parent compound ( Figure 2) were further selected for in silico targetreceptor interaction and ADMET studies.

Binding affinity study through docking against TNF-a
The aim of the molecular docking study was to elucidate whether capsazepine derivatives modulate the anti-inflammatory target and also to identify the binding site against well-known human anti-inflammatory molecular target TNF-a. The native ligand re-docking study indicates that the software predict the reliable results. Thereafter the predicted active derivatives were subjected to molecular docking studies. The docking results provided pertinent information about the binding affinity, binding energy and orientation of ligand-receptor interactions. The docking results are summarized in Table 4. It has been found that capsazepine and its active derivatives bound to the same active site as reported in the PDB protein crystallographic structure database. Capsazepine showed significant binding affinity to TNF-a dimeric structural unit (A and B chain residues) with binding energy of 27.3 kcal/mol, similarly, capsazepine active derivatives CPZ-34, and CPZ-30 showed high binding affinity to TNF-a dimeric structural unit with docking energy of 2 7.8 and 27.9 kcal/mol, respectively. Docking pose of capsazepine and its active derivatives on receptor TNF-a are showed in Figure 5. On the other hand, capsazepine derivatives CPZ-29 and CPZ-35 showed moderate binding affinity to TNF-a dimeric structural unit with binding energy of 27.2 and 27.3 kcal/mol, respectively. The capsazepine derivatives CPZ-33 and CPZ-36 showed low binding affinity to TNF-a dimeric structural unit with binding energy of 26.8 and 26.9 kcal/mol, respectively. The  chemical nature of capsazepine & its derivatives binding site amino acid residues on TNF-a dimeric structural unit (Chain A & B) were aliphatic (e.g., LEU-57, LEU-120, GLY-121, GLY-122), hydroxyl group containing (e.g., SER-60), and aromatic (e.g., TYR-59 and TYR-119). The capsazepine, CPZ-30, CPZ-33 and CPZ-34 showed H-bonds with the TNF-a dimeric unit residues, this suggest high structural stability and may lead to high inhibitory activity of capsazepine and its derivatives on TNF-a active site.

Bioavailability and ADME parameters screening for drug likeness
The compound's good absorption or permeation through blood brain barrier is measure by its LogP that must be less than 5 [222 23]. Results of pharmacokinetic screening revealed that capsazepine and its most active derivatives CPZ-33 and CPZ-34 followed the Lipinki's rule of five for oral bioavailability. However CPZ-29, CPZ-30, and CPZ-36 showed lipophilic nature due to high LogP value, while compound CPZ-35 showed both high lipophilicity and low membrane permeability due to high LogP and molecular weight. These ADMET screening results are summarized in Table 5 and 6. The ADME descriptors of capsazepine and its derivatives were calculated for drug likeness studies. The intestinal absorption and blood brain barrier penetration were predicted by developing an ADME model using descriptors 2D PSA and AlogP98 that include 95% and 99% confidence ellipses. These ellipses define regions where well-absorbed compounds are expected to be found. The results of DS-ADME model screening showed that capsazepine derivatives CPZ-33 and CPZ-34 possess 99% confidence levels for human intestinal absorption and blood brain barrier (BBB) penetration. Similarly, another predicted active capsazepine derivative CPZ-30 also showed 99% confidence level for intestinal absorption and 95% confidence level for BBB penetration. The capsazepine derivatives CPZ-29, CPZ-35 and CPZ-36 fall outside the ADME model ellipses filter, which indicate its poor intestinal absorption and BBB penetration ability. The plot of polar surface area and ALogP fpr capsazepine and its derivatives are represented in Figure 6.

Toxicity risks assessment
The USFDA (US FDA, United States Food and Drug Administration) standard toxicity risk predictor software TOP-KAT (Discovery Studio, Accelrys, USA) locates fragments within the compound that indicate a potential threat to toxicity risk [26]. Toxicity screening results of TOPKAT for capsazepine and its derivatives showed that studied compounds possess no risk of carcinogenicity, mutagenicity and skin irritation, however it possess high developmental or reproductive toxicity potential at high doses or long term therapeutic use in human. The capsazepine and its derivatives CPZ-30, CPZ-33 and CPZ-34 showed strong skin sensitization capacity. Similarly, capsazepine derivatives CPZ-33 and CPZ-34 also showed mild ocular irritancy. Other detail predicted toxicity parameters are summarized in Figure 6. Plot of polar surface area (PSA) versus ALogP for capsazepine and its derivatives showing the 95% and 99% confidence limit ellipses corresponding to the blood brain barrier (BBB) and intestinal absorption. doi:10.1371/journal.pone.0100797.g006 Table 7. A) Compliance of capsazepine and its derivatives to the computational parameters of toxicity risk; B) Compliance of capsazepine and its derivatives to the computational parameters of USFDA rodent carcinogenicity, Ames mutagenicity, developmental toxicity potential, aerobic biodegradability, ocular irritancy and skin irritancy.  Table 6. The results of toxicity risk for capsazepine and its active derivatives showed moderate to good drug score, in compared with capsazepine (Table 7a). Similarly, toxicity screening results of USFDA rodent carcinogenicity, Ames mutagenicity, developmental toxicity potential, aerobic biodegradability, ocular irritancy and skin irritancy also showed positive response to capsazepine and its derivatives (Table 7b).

Anti-inflammatory potential of Capsazepine
To examine the effects of capsazepine on LPS-induced proinflammatory cytokine TNF-a in macrophages, culture supernatant from various treatment groups were used to determine their production. Treatment of capsazepine inhibited (p,0.05) the production of LPS-induced inflammatory mediator in dose dependant manner. (Figure 3, Table 2).

Measurement of the cell viability
The in vitro effect of capsazepine on cell viability in peritoneal macrophage cells isolated from mice was evaluated using MTT assay. The significant change in percent live cell population was not observed (p,0.05) at any concentration of the treatment when compared with normal cells (Figure 4, Table 3).

Conclusions
The experimental in vitro evaluation of capsazepine against proinflammatory mediator TNF-a indicated that capsazepine mediate significant inhibitory effect on the TNF-a. The predicted activity of capsazepine was comparable with the experimental results. Ligand-based virtual screening through developed QSAR model resulted in six best hits for capsazepine derivatives CPZ-29, CPZ-30, CPZ-33, CPZ-34, CPZ-35 and CPZ-36. The capsazepine derivatives CPZ-33 and CPZ-34 showed good predicted activity and binding affinity to TNF-a in compared with capsazepine. Docking results indicate that the major influencing factors of molecular interactions between TNF-a and capsazepine and its derivatives were H-bonds, hydrophobic and electrostatic interactions. Results of oral bioavailability (rule of five), ADME and toxicity risk profiling were within the acceptable limit for capsazepine derivatives CPZ-33 and CPZ-34. These compounds as such and on further lead optimization may guide to designing of novel TNF-a inhibitors.

Supporting Information
File S1 Contains Table S1, Structure, experimental IC 50 (mM), predicted IC 50 (mM) and residual of training set compounds.