Prediction of anticancer property of bowsellic acid derivatives by quantitative structure activity relationship analysis and molecular docking study

Abstract

Context:

Boswellic acid consists of a series of pentacyclic triterpene molecules that are produced by the plant Boswellia serrata. The potential applications of Bowsellic acid for treatment of cancer have been focused here.

Aims:

To predict the property of the bowsellic acid derivatives as anticancer compounds by various computational approaches.

Materials and Methods:

In this work, all total 65 derivatives of bowsellic acids from the PubChem database were considered for the study. After energy minimization of the ligands various types of molecular descriptors were computed and corresponding two-dimensional quantitative structure activity relationship (QSAR) models were obtained by taking Andrews coefficient as the dependent variable.

Statistical Analysis Used:

Different types of comparative analysis were used for QSAR study are multiple linear regression, partial least squares, support vector machines and artificial neural network.

Results:

From the study geometrical descriptors shows the highest correlation coefficient, which indicates the binding factor of the compound. To evaluate the anticancer property molecular docking study of six selected ligands based on Andrews affinity were performed with nuclear factor-kappa protein kinase (Protein Data Bank ID 4G3D), which is an established therapeutic target for cancers. Along with QSAR study and docking result, it was predicted that bowsellic acid can also be treated as a potential anticancer compound.

Conclusions:

Along with QSAR study and docking result, it was predicted that bowsellic acid can also be treated as a potential anticancer compound.

Boswellic acids are usually present in the resin of the plant Boswellia serrata belonging to family Burseraceae, which constitutes about 30% of the whole resin.[1] The chemical constituent of boswellic acids are a pentacyclic triterpene, a carboxyl group and one other functional group and mainly exists in the form of alpha and beta form. The derivatives of bowsellic acid structures may take an additional hydroxyl group at triterpene ring, hence they differs from each other.[2] Ethanol as a suitable solvent is used for extraction of typically boswellic acids from the commercial sources that contain from 37.5% to 65% boswellic acids.[3] Current research has established the fact about the anti-inflammatory actions of the boswellic acids like the conventional nonsteroidal anti-inflammatory drugs (NSAIDs). Acetyl-11-Keto-β-boswellic acid (AKBA) present in Bowsellia extracts are also found to be an inflammatory response by inhibiting 5-lipoxygenase, the enzyme responsible for the biosynthesis of leukotrienes.[4] Recently, also the area enzyme inhibition activity of these compounds also has confirmed by using nuclear magnetic resonance and mass spectrometry and molecular docking analysis.[5] In contrast to classical NSAIDs action that accelerates articular damage in arthritic conditions, boswellic acid significantly reduces the glycosaminoglycan degradation, therefore does not lead to ulcer creation.[6] In addition to anti-inflammatory actions, boswellic acids are also used as much effected for the anticancer, antimicrobial, anti-analgesic, antipyretic, and platelet-inhibitory actions.[7,8] Some other in vitro study found that boswellic acid acetate (BC4), as a potent inducer of differentiation and apoptosis of leukemia cells with 90% of cells showing morphological changes.[9,10,11] Owing to potential action of boswellic acid in therapeutics it is also undergoing an early-stage clinical trial at the Cleveland Clinic.[12,13]

Nuclear factor-kappaB (NF-κB) is usually exists as a heterodimer between Rel and p50 proteins [Figure 1]. While in an inactivated state, NF-κB is located in the cytosol complexed with the inhibitory protein. By a series of action, the extracellular signal activates the enzyme IκB kinase (IKK) which further phosphorylates the IκBα protein which ultimately results ubiquitination, that is, the dissociation of IκBα from NF-κB, and followed by the degradation of IκBα by the proteasome complex. The activated NF-κB is then translocated into the nucleus and the DNA/NF-κB complex helps for transcription of DNA into mRNA, further is translated into protein thereby changing the cell function.[14,15] Similarly, during oncogenesis nuclear NF-κB activity plays an important role in the development and progression of lymphoma, leukemia, and some epithelial cancers. The carcinogenic signals is due to the activation of IkappaB alpha kinase (IKK), which is then activates the NF-κB to mediate the cancer cells for their survival. Thus, inhibition of tumor necrosis factor alpha-induced IKK activity with specific IKK inhibitor represents an interesting strategy to treat cancer.[16]

Figure 1.

Mechanism of nuclear factor kappa-light-chain-enhancer of activated B-cell in signaling action

Theoretically to establish the relationship between molecular property of a molecule and its activity (may be anticancer, anti-arthritis, etc.), quantitative structure activity relationship (QSAR) study is essential, which is required for novel drug design process.[17] Mathematically QSAR models are regression models which link a set of predictor variables to the strength of the response variable. Three main components of QSAR model include, the properties to be modeled, the chemical information and the algorithm/methods used to link the property and activity of the chemical.[18] Similarly, the binding site identification and characterization also the binding affinity of a novel small molecule with its receptor can be obtained by molecular docking methods.[19] Along with QSAR and docking study, in vitro approach has been found suitable for drug property analysis.[20] Phytochemicals from Boswellia resin, that inhibits NF-κB protein activation has been studied in a mouse model for anticancer activity (see the discussion section). Hence hopefully the derivatives of bowsellic acid derivatives are expected to constitute a potential novel group of NF-κB inhibitors. The aim of this study is to predict the property of the bowsellic acid derivatives as anticancer compounds by computational approach. To evaluate this, extensive QSAR study of bowsellic acid derivatives has been performed, followed by molecular docking study to cross verify the result.

Materials and Methods

Retrieval of boswellic acid and its derivatives from PubChem

The structural files of compound boswellic acid, including its analogs were retrieved from PubChem database (http://www.pubchem.ncbi.nlm.nih.gov). The collected data include the structural coordinate file in Inchi format, simplified molecular-input line-entry (SMILE) format, IUPAC name, molecular formula, molecular weight and so on. The main PubChem is a data base released in 2004 that provides much useful information to analyze the biological activities of small molecules. PubChem also provides a fast chemical structure similarity search tool. Further MarvinSketch was used for conversion of two-dimensional file format to corresponding three-dimensional form. Marvin Sketch is sophisticated software that provides an advance and user friendly platform for conversion of molecular format as well as editing and drawing of chemical structures also reactions in a GUI based platform.

Calculation of descriptors from the molecules

The various quantitative features (descriptors) of the molecule derivatives were obtained from preADMET server (preadmet.bmdrc.org) and Molegro tool. The preADMET server provides basically five categories of descriptors.

Quantitative structure activity relationship modeling methods

Regression models make it possible to model and identify relationships in existing data, and to make predictions on unseen data. Minitab 14.0 is used for the regression equation modeling available at http://www.minitab.com/en-us/products/minitab/. For regression analysis, Andrews affinity values were considered as a response (dependent variable) and the computed descriptors were taken as independent variables.[21] Molegro Data Modeler provides four methods for regression analysis, namely multiple linear regression (MLR), partial least squares (PLS) analysis, artificial neural networks (ANN), and support vector machines (SVM). Then, each category of regression models was evaluated by taking the parameters such as Pearson correlation (r), Pearson correlation squared (r²), Spearman Rank Correlation (ρ), mean squared deviation (MSD), root mean squared deviation, cross validated squared correlation coefficient (q²), and best model selected according to the rank.[21]

Docking of ligands (boswellic acid) with receptor protein

In order to study, the potential nature of the derivatives with respect to anticancer property molecular docking was performed. Of 65 derivatives, few molecular structures was chosen based on their computed Andrews affinity value. The receptor, human NF-kappa-beta-inducing kinase structure was obtained from the Protein Data Bank (PDB). The docking simulation was performed by using AutoDock 4.2 tool (autodock.scripps.edu/), that facilitates flexible mode of docking. To the both ligand and receptor polar hydrogens are added and suitable grid was chosen to facilitate the affinity mapping process. The Lamarckian genetic algorithm was used for the docking process.

Results

Retrieval of boswellic acid derivatives from PubChem and computation of descriptors

The SMILE format of 65 numbers of boswellic acid (CID 168928) and its analogs was retrieved from the PubChem database. All were converted to their corresponding two-dimensional structural format as well as a Mol2 format by using a Marvin Sketch tool (http://www.chemaxon.com/products/marvin/marvinsketch/). Molecules with a molecular weight <500 (g/Mol) were selected from the PubChem database (supplementary material 1). The generated Mol2 files were exported to Molegro Virtual Docker followed by calculation wizard further pipelined to Molegro data modeler. From the result the Andrews affinity descriptor was selected for the dependent variable in the modeling process. Furthermore, all 65 ligands were imported to preADMET work space and five respective categories of descriptors were calculated. The five classes of two-dimensional descriptors considered for calculation are Constitutional, Electrostatic, Geometrical, Physicochemical, and topological type (supplementary material 2).

Quantitative structure activity relationship analysis of all categories of descriptors

The descriptors and the dependent variables were correlated by regression analysis by using Minitab-14.[22] All total eight numbers of the response variable were taken as independent variable for regression analysis as given in following regression equations.

Andrews (Constitutional) = −13.9 − 0.443 No-H + 2.80 No-O − 0.844 H-Bond Donor + 6.79 No-Ring + 0.078 No-Rigid Bond + 0.031 No-Single Bond − 1.54 No-Double Bond (Equation 1)

Andrews (Electrostatic) = −6.2 + 0.393 PPSA1 − 0.696 PNSA1 − 0.724 WPSA1 + 1.52 WNSA1 − 0.0387 hPCS + 0.0784 hNCS − 1.37 Q-max + 0.26 Q-min (Equation 2)

Andrews (Geometrical) = −12.3 − 0.0880 hA − 0.00404 PSA + 0.0111 HBASA + 0.0277 HBDSA − 0.0432 t- PSA + 0.0688 2VWSA + 0.0716 2VWSV+0.0779 0GSA (Equation 3)

Andrews (Physicochemical) = −13.7 + 0.318 Pol − 2.17 Sklogp + 0.000047 WSBS + 2.71 Alogp98 − 0.355 WSFE + 0.073 WSPW − 0.00411 SKBP + 0.0203 SKMP (Equation 4)

Andrews (Topological) = 11.0 − 0.000414 WI + 0.000000 2HI + 0.295 ZI + 0.170 QI + 0.0101 R − 0.0830 ECI − 1.45 EAI − 0.00112 EMTI (Equation 5)

Further various statistical parameters were computed for the validation purpose by using Molegro tool. The results were obtained from different methods such as MLR, PLS, ANN, and SVM was analyzed comparatively [Table 1].

Table 1.

Molegro output of comparative statistics among various descriptors

Among all the methods of QSAR analysis the ANN based method that constitutes a geometrical type of two-dimensional descriptors show the highest Pearson correlation (r) =0.987 with less MSD = 0.70608 that indicates that the geometrical type of descriptors is very much significant for the bio-activity of the bowsellic acid molecules. This result also a proof for the ANN method which is a promising approach for the best statistical approximation and would be helping in solving complex problems.

Docking study

The bio-activity of bowsellic acid derivatives was assumed to be based on Andrews affinity value, as it is related to ligand binding affinity to the receptor.[23] The ligands were selected based on binding affinity of higher values for Andrews affinity descriptor (supplementary material 2). Best six numbers of bowsellic acid analogs 33, 34, 51, 57, 59, 63 were selected and were undergone further energy minimization by the PRODRG server (http://davapc1.bioch.dundee.ac.uk/prodrg). Similarly the receptor human NF-kappa-beta-inducing kinase crystal structure was obtained from the (PDB ID 4G3D).[24] Docking was performed between NF-κB with above stated best selected analogs of boswellic acid using AutoDock 4.2 tool and the orientation of ligand and binding energy was computed [Figure 2].

Figure 2.

Showing Docking view of receptor (in wire frame) and ligands (analog 33-red, analog-34-green, analog 51-blue, analog 57-yellow, analog 59-pink, analog 63-cyan color) that makes hydrogen bonds (green dotted lines)

From this docking result, it was obtained that Analog-57 (3β, 6β, 17α, 18ξ)-3, 6, 16, 23-Tetrahydroxyolean-12-en-28-oic acid) having the highest binding energy −9.06. The bind pocket was visualized by Molegro Virtual Docker and residues in the binding pocket include Gly-592, Gln-349, Gly348, Leu-382, Leu-455, Pro-454, Cys-444, Gly 446, Leu-447, Thr-448 also having the highest Andrews value 17.809 [Figure 3]. This indicates that, the Andrews affinity is the best response to calculate the biological activity of the drug molecule by using computational method.

Figure 3.

(a) Structure of analog 57 (b) residues of protein in blue color and analog 57 in the cavity (yellow color), the three hydrogen bonds are shown in green dotted lines

Discussion

According to an in vitro analysis report, Bowsellia resin contains incensole acetate that inhibits IκB kinase (IKK) activation loop and the kinase activation does not affect NF-kB pathways.[25] However, subsequent studies also showed that AKBA could strongly inhibit tumor angiogenesis thereby down regulation of cancer-associated biomarkers.[26,27,28] Boswellic acid derivatives also exhibit a range of cytotoxicity against various human cancer cell lines, hence attempts to identify a potential lead compound as an inhibitor of the NF-κB and STAT proteins have been performed by QSAR analysis.[29] Recently, ANN/modified ANN method are used mostly in QSAR analysis.[30] The result obtained in our study indicates that, the ANN based model is fitted as the best correlation pattern from our comparative analysis of different QSAR models, is in a good agreement with this. Similarly, geometrical descriptors of drug molecules are important features that relate to the binding thereby enhancing its activity as these are related to three-dimensional structure.[31] Molecular docking is used to evaluate the binding affinity as well as to identify the pattern of binding. Recently, docking study has been performed on NF-kB as taking as a receptor to evaluate the phytochemicals as the breast cancer inhibitor.[32]

Conclusion

In this work, a computational approach has been used to evaluate the anti-cancerous nature of bowsellic acid derivative. The derivative 3β, 6β, 17α, 18ξ)-3, 6, 16, 23-Tetrahydroxyolean-12-en-28-oic acid was obtained as the potential NF-kappa-beta-inducing kinase inhibitors in the major signaling NF-κB pathway. Hence, it might stand for a substitute drug for classical medicine treatments for cancer. Along with Insilico based QSAR study, molecular docking is an important method to correlate the chemical structure to the bio-activity prediction.