Computational insights into the identification of a potent matrix metalloproteinase inhibitor from Indigofera aspalathoides to control cancer metastasis

Abstract

Matrix metalloproteinases (MMPs) are the major proteolytic enzymes which assist in regulating the metastatic process by degrading the extracellular matrix proteins. In this study, we have investigated the anti-metastatic potential of major bioactive compounds in the medicinal plant Indigofera aspalathoides targeting matrix metalloproteinases (MMP2 & MMP9) and it’s in silico pharmacokinetic profiles using computational studies. Indigofera aspalathoides (Sivanar vembu in Tamil) is a renowned medicinal herb in traditional Indian medicine which contains indigocarpan, mucronulatol, indigocarpan diacetate, erythroxydiol X and erythroxydiol Y as the major constituents. The 3-dimensional structure of MMP2 and MMP9 was designed by using I-tasser and Modeller and it was validated by PROCHECK. The structures of mucronulatol and indigocarpan have been retrieved from PubChem and indigocarpan diacetate, erythroxydiol X & Y were drawn by using Chemdraw Ultra 6.0. Batimastat was used as a positive control. Molecular docking was performed by using AutoDock 4.2 tools and AutoDock vina, an open-source program which signifies an effective interaction between the phytoligands and MMP2 & MMP9. From the results, AutoDock 4.2 have showed that indigocarpan possesses strong binding energy (ΔG) of − 7.68 kcal/mol towards MMP2 and − 6.35 kcal/mol towards MMP9, whereas batimastat showed binding energy (ΔG) of − 6.34 kcal/mol for MMP2 and − 5.66 kcal/mol for MMP9, meanwhile the results from AutoDock vina indicates that indigocarpan possesses strong binding energy (ΔG) of − 8.0 kcal/mol towards MMP2 and − 8.2 kcal/mol towards MMP9, whereas batimastat showed binding energy (ΔG) of − 7.2 kcal/mol for MMP2 and − 7.6 kcal/mol for MMP9. Also, the ADME and toxicity results suggest that the indigocarpan compound possesses a druggable pharmacokinetic potentiality and does not have carcinogenicity and Ames mutagenesis compared with other phytoligands. Hence, it is evident from our results that both AutoDock platforms strongly revealed that the phytoligand, indigocarpan possesses strong inhibitory activity against MMP2 and MMP9 to control cancer metastasis.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13205-021-02731-w.

Introduction

Cancer metastasis is a complex mechanism in which a single tumor cell or a mass of cells propagate from a primary tumor site to the secondary distant organs and tissues and it is the fundamental rationale for cancer mortality (Winer et al. 2018). Tumor cells execute metastatic potentiality once it attains gainful characteristics, which relocates from the initial site, thereby it penetrates the blood vasculature and enters the secondary sites and it forms metastatic foci (Loffek et al. 2011; Cathcart et al. 2015). Statistically, metastasis is accountable for almost 90% of cancer-related deaths (Kumar et al. 2018). It is a complicated mechanism that encounters a sequence of systemic events known as invasion-metastasis cascade (Seyfried et al. 2013; Fares et al. 2020). This cascade involves five different steps: (a) invasion and migration—cancer cells dissociate from the primary tumor and invade neighboring, normal tissues. (b) intravasation—the incursion of tumor cells into the blood capillaries in which the tumor cells secrete proteolytic enzymes (matrix metalloproteinases) that allow them to permeate the blood vessels. (c) circulation—the abnormal cells which pass through the bloodstream and it has to confront certain conditions such as high oxygen and cytotoxic lymphocytes. (d) extravasation—cells frequently get adhered into the tissues which leave blood capillaries by infiltration of the endothelium through cell proliferation. (e) Colonization—tumor cells that habitats at a distant tissue site form a secondary tumor. Thereafter, it undergoes cell proliferation and stimulates neoangiogenesis to establish adequate vascularization (Deryugina et al. 2006; Kessenbrock et al. 2010; Kunz et al. 2016).

The presence of proteolytic enzymes plays a crucial role in tumor metastasis, i.e. Matrix metalloproteinases (MMPs) (Quintero-Fabián et al. 2019). It is also known as matrixins, is a Ca²⁺ and Zn²⁺ dependent protein which is classified into six families based on its substrate specificity (Leber et al. 2009). The interactivity of cells among the ECM is essential for the pathological modulation that exists during carcinogenesis. A certain number of ECM proteins influence the phenotype of cancer cells and their outcome during cell migration. MMPs were usually involved with assisting metastasis by breaking the ECM barriers (Rathee et al. 2018). Though MMPs are joined with tumor cell’s survival and growth, they are generated by the cancer cells in a few amounts. Cancer cells secrete various cytokines and growth factors that provoke nearby host cells to synthesize required MMPs. The secreted MMPs get bounded onto the tumor cells surface and thereby, it is involved in enhancing the different stages of carcinogenesis (Jabłonska-Trypuc et al. 2016). Among the MMPs, MMP-2, and MMP-9 (Gelatinase) possesses a significant role in tumor development and angiogenesis in breast cancer and prostate cancer (Zhou et al. 2014; Woo et al. 2017; Parhi et al. 2017). In the current study, we aimed to identify the anti-metastatic drug therapy from the medicinal plant, Indigofera aspalathoides Vahl using computational prediction. I. aspalathoides belong to the family Papilionaceae and it is used for the treatment of various disorders in the Traditional Indian System (Sivagnanam et al. 2012). It contains pterocarpan derivatives, saponins, tannins, steroids, alkaloids, flavonoids, and reducing sugars as essential constituents (Swarnalatha et al. 2015). The plant has anti-tumor activity (Rajkapoor et al. 2004), COX inhibitory activity (Selvam et al. 2004) and anti-inflammatory activity (Bhagavan et al. 2013). The compound possessing the anti-metastatic property remains unknown. The major bioactive constituents are indigocarpan, mucronulatol, indigocarpan diacetate, erythroxydiol X, and erythroxydiol Y (Saraswathy et al. 2013). Thus, finding new anti-metastatic drugs from the plant-derived biomolecules could be a better intention to control metastasis by intruding matrix metalloproteinases. Though many phytochemical investigations have been reported in I. aspalathoides, the active biomolecule to control metastasis has not been yet investigated. In this perspective, the present study has been designed to evaluate the anti-metastatic potential of active biomolecules of I. aspalathoides with MMP-2 and MMP-9 and it’s in silico pharmacokinetic principles using computational methods.

Materials and methods

All in silico computational predictions have been performed in PC Windows 7 using the Intel Core™ i5 processor, 4 GB RAM, and its 32bit operating system.

Softwares used

The Uniprot has been used for retrieving the FASTA sequences of the target protein and the protein model was built by I-Tasser and it was cleansed by GalaxyRefine. The protein 3D construct was evaluated using ERRAT and PROCHECK. The three-dimensional structure of plant-derived ligand molecules was obtained from PubChem. Online tools such as SwissADME and admetSAR have been used to study the pharmacokinetics and toxicity profiles of the phytoligands. Molecular docking has been accomplished by AutoDock 4.2 program and the interactions between the target protein and the ligand molecules were visualized by Accelrys Discovery Studio Visualizer 2.5 (Accelrys Software Inc., San Diego, CA).

Phytoligands preparation

3D structure of the ligand molecules indigocarpan (https://pubchem.ncbi.nlm.nih.gov/compound/10245208), mucronulatol https://pubchem.ncbi.nlm.nih.gov/compound/442811) and batimastat (https://pubchem.ncbi.nlm.nih.gov/compound/5362422) was obtained from PubChem database. Batimastat was used as a positive control. Other ligand molecules indigocarpan diacetate, erythroxydiol X, and erythroxydiol Y were drawn by using Chemdraw ultra 6.0. All the ligands were prepared by using AutoDock 4.2 and their fidelity was examined by using “Add hydrogen” options. Lastly, all compounds were stored in PDB file format and it was availed for docking and toxicity predictions.

Protein preparation

Based on the reports from the previous literature, Matrix metalloproteinases (MMP2 & MMP9) have been selected as the most probable metastatic drug target for the plant-derived small molecules.

SOPMA: prediction of secondary structure

MMP2 (Uniprot ID: P08253) and MMP9 (Uniprot ID: P14780) secondary structure were resolved by using the Self-Optimized Prediction Method with Alignment (SOPMA). The program predominantly estimates the presence of amino acid residues of about 69.5% which is deliberative in depicting protein secondary structure (α-helix, β-sheet, and coil) concerning the MMPs within a complete database comprising of 126 chains of non-homologous (< 25% similarity) proteins (Geourjon et al. 1995).

Molecular modeling of matrix metalloproteinases

Retrieval of FASTA sequences

The whole protein sequence concerning human matrix metalloproteinases (MMP2 & MMP9) which contains 660 (MMP2 Uniprot ID: P08253) and 707 (MMP9 Uniprot ID: P14780) amino acid residues were retrieved from Uniprot.

Construction of 3D model using I-Tasser and Modeller 9.25

Homology modeling was performed to resolve the 3-dimensional structure of human matrix metalloproteinases (MMP2 & MMP9) using I-TASSER (https://zhanglab.ccmb.med.umich.edu/I-TASSER/). The I-Tasser server is an online platform to predict the structural and functional annotation of the target macromolecule using an automated procedure (Yang et al. 2015). This program enables the user to construct the 3-dimensional structure of the protein that emulates sequence to structure to the function module. It retrieves the template protein from the PBD library by using multiple threading alignment strategies (LOMETS). The fragments congregate into a complete framework through the Monte Carlo simulation method, also the unaligned regions especially the loops and tails were built through the ab initio modeling. Spatial restraints enhance the simulation process to select the appropriate model from the clustering decoys based on its lowest energy. Finally, the hydrogen bonding optimization enables to obtain the best model based on its function scores (C-score). I-Tasser represents the C-score as a confidence score which is aimed at manipulating the importance of the assessed models (Yang et al. 2015). Greater the C-score value (between − 5 and 2), implies that the predicted model procures higher confidence. The 3D structures of human MMPs were designed by using Modeller 9.25. The FASTA sequences of human MMPs (MMP2 and MMP9) were retrieved from the Uniprot database. The obtained sequences were utilized for BLASTp analysis against the protein data bank (PDB) to identify the 3D protein structure based on multiple sequence alignment (MSA). The human Gelatinase A (PDB ID: 1CK7), Catalytic domain of proMMP2 (PDB ID: 1EAK) and Crystal structure of human matrix metalloproteinase 9 (PDB ID: 1L6J) were used as the template structures. By using the templates, 3D structural models were generated and validated using the Modeller (v9.25) (Marti-Renom et al. 2000). Based on DOPE (Discrete Optimized Protein Energy) score functions, the models were ranked by the DOPE statistical potential and the best model with lowest DOPE score was assessed for the stereochemical properties and to determine the Ramachandran plot using PROCHECK online program.

Model refinement

The refinement process has turned out to be the last step of the protein structure prediction journey to attain equivalence with experimental accuracy. Among the various models obtained from the I-TASSER server, the topmost construct was selected and proposed for the refining process of the whole 3D model to the GalaxyRefine server (http://galaxy.seoklab.org/). GalaxyRefine predicts the whole protein model accurately from the pattern using template-based modeling and it is a reliable method (Shin et al. 2014). The refinement process was performed which is based on CASP10 that emulates side-chain reconstruction and molecular dynamics relaxations. In compliance with CASP10, the GalaxyRefine was found to be a foremost algorithm that strengthens the structural packing and thus ameliorates the structural features of the side-chain and also the backbone of the protein model. Initially, the side chain arrangements were refurbished using rotamers and the CHARMM22 force field determines their energy functions. To determine the accuracy of protein quality with initial models, it could be calculated using high-accuracy global distance test (GDT-HA), side-chain global distance test (GDC-SC) for evaluating the accuracy of local structure, and MolProbity scores were estimated for physical correctness of the model (Heo et al. 2013).

Validation of the 3D model

Protein validation enables to classify plausible misconception in the refined models. Refined 3-dimensional protein structures have been validated by using distinct online programs viz. ERRAT and PROCHECK. For analyzing the protein structure specifically about the energy and its stereochemical geometry, the PROCHECK program was used. It works by resolving residue by residue geometry and it was subjected towards a pattern of analysis to determine its reliability and internal consistency. The presence of the backbone was confirmed and examined through assessing the psi/phi Ramachandran plot procured using PROCHECK (Laskowski et al. 1993). ERRAT (http://services.mbi.ucla.edu/ERRAT/) was known to be as the overall quality factor which appraises the constructed 3D construct through a notable crystal structure and assess the error values, by enumerating the non-bonded atomic interactions. Based on the acquired errors, the overall quality factors could be calculated through high-resolution structures (Messaoudi et al. 2013). Depending on the variation in consensus results from all models achieved through the servers, topmost 3D structures were preferred for docking studies. From the selected model the Zn²⁺ ions were introduced. In MMP2, Zn²⁺ was introduced by combining the modeled MMP2 structure with human Gelatinase A (PDB ID: 1CK7) and similarly, the Zn²⁺ ion was introduced in MMP9 by combining the modeled MMP9 with the crystal structure of human MMP9 (PDB ID: 1L6J). The parameters such as van der Waals radius σ = 1.10 Å and the Lennard Jones potential well ε = 0.0125 kcal/mol were fixed for Zn²⁺ ions. It was then visualized by using Pymol (Boopathi et al. 2020; Moreira et al. 2020).

Ligand-binding site prediction

For locating the ligand-binding regions in MMP2 and MMP9, COACH (https://zhanglab.ccmb.med.umich.edu/COACH/) online program was employed. It is a meta server application that is a dual process consisting of TM-SITE and S-SITE, was used to identify the most active sites in the protein (Yang et al. 2013). To obtain a consequent ligand binding site, S-SITE and comparative TM-SITE computed values were merged towards COFACTOR, FindSite, and ConCavity. This server predicts a discernible ligand-binding residue, where the respective ligand can bind protein macromolecule, collectively by calculating EC (enzyme commission number) and GO (gene ontology) fundamentals through favoring the protein macromolecule (Yang et al. 2013).

F pocket

It is an open source platform for the identification of active pockets in the protein macromolecule. It is based on voronoi tessellation and the presence of alpha spheres built on the package. It is available under http://fpocket.sourceforge.net. It has been written in C-programming language which is well suited to develop and extract active pocket descriptors for our protein of interest (Guilloux et al. 2009).

Molecular docking simulation: AutoDock 4.2

Molecular docking was performed for the selected phytoligands with the 3D structure of MMP2 and MMP9 using AutoDock Tools 4.2 program (Morris et al. 2009). Initially, the 3D structure of MMP2 and MMP9 were designed using I-TASSER and they were progressed by the certain procedure to be followed in the preparation of protein molecule that includes: (i) integration of hydrogen atoms, (ii) incorporation of less polar hydrogen molecules, (iii) distribution of Kollman charges and (iv) modifying it as PDBQT files. The 3-dimensional structure of selected phytoligands (Fig. 1) have been retrieved from the PubChem database, and it has been followed by (1) addition of Gasteiger’s charges, (2) integration of polar hydrogens, (3) incorporation of non-polar hydrogen molecules, (4) establishing rotatable bonds and (5) saving the file as PDBQT format. Following the subsequent course of action in formulating target macromolecule and ligand, their preparation was preceded by Molecular Graphics Laboratory (MGL) Tools. Furthermore, docking studies have been carried out by assuming target protein as rigid whereas ligand molecule is a flexible molecule. Accordingly, the generation of grid map has been formed for the receptor through which it focuses mostly on the BH3 domain via auto grid program (X = 48 Å, Y = 48 Å, Z = 48 Å, grid spacing value of 0.375 Å and its center point X = 83.69 Å, Y = 85.24 Å, Z = 87.057 Å). During computation, the Lamarckian genetic algorithm (LGA) was compiled as 100 GA runs, population-level is 150, the mutational range is 0.02, and their crossover value was found to be 0.8. Both AutoDock 4.2 and AutoGrid 4.2 computational analysis was achieved to generate grid parameter file (gpf), also docking parameter file (dpf) that are highly specific for the ligands. Based on this binding energy between protein and the phytoligand, different forms of conformations could be classified into a cluster form, and also foremost docked conformation was observed. The lower binding energy concerning the protein–ligand complex could be appraised as the prominent binding orientation. Ultimately, the consequence of the docking result was directly dragged into Discovery Studio Visualizer for finding varying interactions between the protein–ligand complexes.

Fig.  1.

Structure of phytoligands and batimastat (positive control)

AutoDock Vina

AutoDock vina is an open-source program written in C +  + which runs in all operating systems. The inhibitory potential of phytoligands from I. aspalathoides on MMPs was studied by docking experiments using AutoDock vina program (Trott et al. 2010). The 3D structures of MMPs (MMP2 and MMP9) were constructed by modeller and MMP2 contains 660 residues and MMP9 contains 707 residues. The water molecules were removed and polar hydrogen atoms were added to both MMPs followed by introducing the Kollman charges, it was then saved as Pdbqt file. Molecular docking was performed through AutoDock vina programming by focussing the major active pockets in both MMPs. A grid box was introduced for MMP2 and MMP9 protein to focus specifically on the active pockets of both MMPs. The grid box on the active region of MMP2 was performed at grid center X = 60.81 Y = 86.07 Z = 152.413 with dimension 48 × 48 × 48 (X × Y × Z) and spacing 1 Å similarly, grid box on the active region of MMP9 was performed at grid center X = 32.377 Y = 52.900 Z = 10.069 with dimension 40 × 40 × 40 (X × Y × Z) and spacing 1 Å. Finally, the protein–ligand complex was visualized by using Discovery Studio Visualizer.

ADME and toxicity prediction

The physicochemical influence for phytoligands has been employed to evaluate the pharmacokinetic attributes of different phytoligands within a biological environment. SwissADME and admetSAR were used to study the in silico ADME and the toxicity features concerning various phytoligands from I. aspalathoides (Paramashivam et al. 2015; Daina et al. 2017; Han et al. 2019). The absorption capability of phytoligands depends mainly on membrane permeability, human intestinal absorption, skin permeability, and Pgp. Some of the features influences scattering of phytoligands that comprise the BBB, volume of distribution, and permeability. The metabolism process depends primarily on cytochrome substrates (CYP2C19, CYP2C9, CYP2D6, CYP3A4). Besides, the toxicity of phytoligands was examined by AMES mutagenicity, hERG inhibition, and carcinogenicity. To resolve the in silico ADME and toxicity, the smile format concerning various phytoligands submitted onto the SwissADME (http://www.swissadme.ch/) and admetSAR online server (http://lmmd.ecust.edu.cn/admetsar2/). SwissADME online server analyzes different parameters, such as physicochemical properties, lipophilicity, water-solubility, pharmacokinetics such as GI absorption, BBB penetration, and drug-likeness. Similarly, the admetSAR server predicts toxicity factors such as mutagenicity, carcinogenicity, and hERG inhibition for all the ligands respectively.

Results and discussion

Protein secondary structure prediction

The major intrinsic factors associated with the protein secondary structures are helix, strands, and coil. Depending on its secondary structure, the tertiary structure of the protein could be validated. In our study, we have used SOPMA to determine the secondary structure of MMP2 and MMP9. From results, MMP2 secondary structure comprises beta-turn (T) of 10.61%, extended strand (E) of 21.82%, random coil (C) of 47.12%, and α-helix (H) of 20.45% elements (Fig. 2a). Similarly, MMP9 constitutes beta-turn (T) of 8.20%, extended strand (E) of 20.08%, random coil (C) of 53.89%, and α-helix (H) of 17.82% elements (Fig. 2b).

Fig.  2.

a SOPMA – Secondary structure prediction of MMP2. (blue—alpha-helix, red—extended strand, green—beta-turn, yellow—random coil). b SOPMA – Secondary structure prediction of MMP9. (blue—alpha-helix, red—extended strand, green—beta-turn, yellow—random coil)

Homology modeling of matrix metalloproteinases – I-Tasser

The 3-Dimensional structure of MMP2 and MMP9 was constructed by using I-Tasser, and it follows two distinct approaches. The first approach includes the template retrieved from the PDB library for the 3D prediction which is proMMP2/TIMP2 complex (1GXD) for both MMP2 and MMP9. The sequence similarity across the MMP2 and template was 0.998 and between MMP9 and the template was 0.483. The second approach was resolved by using the ab initio model, in which the construction of MMPs 3-dimensional structures was based on their energy functions. The pattern related to the precise domain of MMP2 and MMP9 has been threaded by the 1GXD (proMMP2/TIMP2 complex) template which was retrieved from the protein data bank (PDB) by LOMETS. Individual fragments were dragged from the aligned regions of the template and it has been constructed into a whole 3D model by REMC (Reaction Ensemble Monte Carlo) simulation mechanism. For the second cycle of the simulation process, spatial restraint extracts have been utilized by REMC. Re-assembling of this 3D model was further refined by FG-MD for creating the final 3D construct. I-Tasser evaluates 3D structure through two different parameters, as C-score and TM-score. Normally, C-score occurs between −5 to 2, whereas the C-score of a higher value implies that a specific 3D construct has high confidence. From the results, the C-score for MMP2 was found to be 0.86 and for MMP9 is − 0.22. It implies that both models are found to be significant with high confidence values. TM-score (template modeling score) is a notable specification which is considered mainly for enumerating those structural similarities between two molecules. The value of > 0.5 implies that the 3-dimensional construct of protein possesses factual topology, whereas the value of < 0.17 implies the presence of an unusual type of structural identity. The TM-score for MMP2 was 0.83, and for MMP9 was found to be 0.68. The configuration of protein structure using normalized Z-score > 1 implies that it could be a good alignment in the 3D models. The normalized Z-score corresponding to the threading templates for MMP2 was 3.54 and MMP9 was found to be 3.13. Among the proposed disparate models, the finest protein models have been refined by using the GalaxyRefine server which emulates the functions of the Critical Assessment of Techniques for Protein Structure Prediction (CASP) method (Lee et al. 2016).

Protein validation

The 3-dimensional structures of the MMPs were validated by ERRAT and PROCHECK programs. PROCHECK evaluates the stereochemical properties of the designed MMPs by perceiving its residues geometrical functions. Results from PROCHECK indicate that the Ramachandran plot of the modeled protein molecule procures the distribution of residues in a distinct environment in both MMPs. In MMP2, among the 660 residues, 473 residues (86.3%) was spotted in the most favored region, 63 residues (11.5%) were observed in the additional allowed region, 6 residues (1.1%) were placed in the generously allowed region, and 6 residues (1.1%) in the disallowed region. Similarly, in MMP9 which constitutes 707 residues, 493 residues (85.1%) was observed in the most favored region, 66 residues (11.4%) were identified in the additional allowed region, 9 residues (1.6%) were placed in the generously allowed region and 11 residues (1.9%) in the disallowed region. Differences in phi–psi dihedral angles intended for every residue between the models were shown in the figure. Dark shading was observed in Chi1–Chi2 plots which show the favorable conformations of residues. Several specifications for main chain parameters were illustrated, for example, Ramachandran plot, hydrogen bonding energy, carbon tetrahedral deformity, poor nonbonding interactions, peptide bonding proportionality, and overall G-factor value. Consequently, the various parameters that stated-above reflect the quality of the modeled 3-dimensional structure. Accordingly, the outcome of PROCHECK implies the consistency of the selected 3-dimensional protein structure. ERRAT delineates the interactivity between diverse molecules, and the outcome of a range greater than 50 is commonly acceptable (Kaushik et al. 2019). In MMPs, the ERRAT score was found to be 81.66 for MMP2 (Fig. 3a) and 74.06 for the MMP9 (Fig. 3b) model which is an acceptable value for a typical 3Dmodels.

Fig. 3.

a Protein model validation by ERRAT for MMP2. b Protein model validation by ERRAT for MMP9

Modeller 9.25

As homology modeling was found to be the backbone of structural biology, the 3D structure of the protein of interest can be built using different machine learning algorithms. It is a more consistent and widely preferred technique for designing unknown protein molecules (Webb et al. 2016). The experimental structures of MMP2 and MMP9 are found to be limited in PDB. The templates for each query sequence were performed using BLAST search. According to the BLAST search of MMP2, it reflects 99.84% template coverage with the Gelatinase A (PDB ID 1CK7) with a resolution of 2.80 Å and 99.76% identity with the catalytic domain of prommp2 (PDB ID 1EAK) with a resolution of 2.66 Å. Similarly, the BLAST search for MMP9 shows 99.76% template coverage with Gelatinase B (PDB ID 1L6J) with a resolution of 2.50 Å. So, the human matrix metalloproteinases MMP2 and MMP9 models were developed by using the MODELLER tool. The internal metrics of the model built using MODELLER are the Molpdf functions which indicate the total energy depending on the model residues. Meanwhile, DOPE (Discrete Optimized Protein Energy) functions also based on the statistical capability of optimal strength obtained from the calculations between the probable distribution function of the given protein and the atomic distance of their native structure. The function of GA341 is calculated from the standard score (Z score) which is a measurement of structural compactness, ranges from 0 to 1, where 0 reflects a poor quality and 1 indicates the model comparable to their native structures. From our results, among the different models, the best model which scored the least DOPE score (MMP2: − 69,060.08594; MMP9: − 55,108.35156) as per the MODELLER scoring functions has been chosen as the final model and both MMP2 and MMP9 possess Z score of 1 and this model has been further validated by using PROCHECK.This tool examines the residue by residue stereochemical property, geometrical functions and scattering of Phi and Psi angle amino acids in favoured, allowed and generously allowed region of a modelled proteins. The Ramachandran plot showed 91.6% amino acids in the core region or most favoured regions and 7.8% residues were in the additional allowed regions and similarly, Ramachandran plot showed 82.7% amino acids in the core region or most favoured regions and 13.1% residues were in the additional allowed regions. The presence of most amino acids residues in the favoured region signifies that the modelled structure of the both MMP2 and MMP9 is reliable, and it can be equally compared with NMR structure quality (Fig. 4a, b).

Fig. 4.

a 3-Dimensional structure of MMP2 with Zn²⁺ ions. b 3-Dimensional structure of MMP9 with Zn^2 + ions

Prediction of active regions

The PDB files corresponding to the MMP models have been bestowed into the COACH server to identify the ligand-binding sites by differentiating a template pattern in the BioLip database. Almost ten distinct models were examined by COACH, and a top model has been chosen which is on the basis of C-score. The reliability of the protein primarily depends on the higher C-score value of the protein. All the predicted residues are mainly accountable for forming active site pockets for the phytoligands. For MMP2, the foremost PDB hit, 1su3B, was preferred and it comprises the C-score of 0.31 and a cluster size of 16 (Table 1). Similarly, for MMP9, PDB hit, 1su3B was chosen and it has a C-score of 0.23 and a cluster size of 15 (Table 2). The outcome from the COACH server is the combined results of algorithms, obtained altogether with COFACTOR, TM-SITE, S-SITE, FindSite, and ConCavity.

Table 1.

Active site prediction for MMP2–COACH Server

S. no.	C-Score	Cluster size	PDB hit	Consensus binding rßesidues
1	0.31	16	1su3B	167,168,170,200,201,202,204
2	0.30	12	1ck7A	102,403,407,413
3	0.18	9	1I6jA	185,186,187,188,189,190,208,211
4	0.12	6	1eakA	130,131,133,209,276,277,278,293,303,304
5	0.08	4	1su3A	476,521,569,618
6	0.07	4	1eakC	296,299,301
7	0.05	3	1fbIA	178,180,193,206
8	0.04	3	4auoB	134,175,209,210,211
9	0.01	1	–	524,529,533,535,542,556,559,591,596
10	0.01	1	1FBLA	56,100,101,102,103,104,185,188,189,190,191,192,193,211,395,400,404,423

Table 2.

Active site prediction for MMP9–COACH Server

S. no.	C-Score	Cluster Size	PDB Hit	Consensus Binding Residues
1	0.23	15	1su3B	164,165,167,197,198,199,201
2	0.20	11	4auoA	182,183,184,185,186,187,205,208
3	0.18	8	1gxdA	99,101,401,405,411
4	0.08	7	4g0dA	114,179,185,186,187,188,189,190,191,192, 193,198,393,401,410,411,420,421,422,423
5	0.08	5	4fu4A	175,177,190,203
6	0.05	4	1I6JA	131,206,208
7	0.02	2	1EAKD	274,275,276,290,300,301,332,334,343,345
8	0.02	2	1eakB	352,354,355,356,357
9	0.02	2	1SU3B	101,102,105,106,107,108,179,190,191,192
10	0.01	1	1EAKB	296,297,298

F pocket

To perform the molecular docking experiment, the active regions in both MMP2 and MMP9 were determined using fpocket. This algorithm involves three main steps. Initially, fpocket prefilters the spheres from the whole ensemble which was determined from the protein structures. The second step helps in recognizing the cluster of spheres to detect active pockets and also to eliminate poor clusters. The final step evaluates the properties from the atoms of the pocket to score each pocket. Various amino acid residues lining the active pockets have been identified using fpocket. The active residues of MMP2 are Phe80, Asn92, Thr96, Lys99, Ala192, Leu399, Gly415, Ala419, Leu420, Ile424, Thr426, Phe431, Val464, Arg495, Met504, Pro506, and Phe512 (Fig. 5a). Similarly, the active residues of MMP9 are Leu28, Val29, Leu30, Phe31, Leu407, Leu409, Phe521, Gly545, Arg546, Gly547 and Ser548 (Fig. 5b). Through, fpocket server various pockets were generated with fpocket score and druggability score (Table 3). Among the different pockets, the top scoring pocket has been chosen for molecular docking evaluation.

Fig. 5.

a Identification of active pocket in MMP2 using F-Pocket open source. b Identification of active pocket in MMP9 using F Pocket open source

Table 3.

FPocket – Active site recognition for MMP2 and MMP9

S. no.	Target macromolecule	Consensus binding residues
1	Human MMP2	PHE80,ASN92,THR96,LYS99,ALA192,LEU399, LEU420,ILE424,GLN415,ALA419,THR426, PHE431,VAL464,ARG495,MET504,PRO506, PHE512
2	Human MMP9	LEU28,VAL29,LEU30,PHE31,LEU407,LEU409, PHE521,GLY545,ARG546,GLY547,SER548

Molecular docking simulation

To study the highest conformational pose of the phytoligands, i.e. indigocarpan, mucronulatol, indigocarpan diacetate, erythroxydiol X, erythroxydiol Y, and batimastat (positive control), molecular docking has been availed using AutoDock (version 4.2) software through the Lamarckian genetic algorithm to investigate the binding affinities, orientation, and the type of interaction between the phytoligands with MMPs. Different parameters that have been evaluated using AutoDock were binding energy, internal energy, intermolecular energy, and torsional energy. Among the phytoligands, indigocarpan binds more effectively to the MMPs by the calculated binding free energy (∆G) of − 7.68 kcal/mol for MMP2 and − 6.35 kcal/mol for MMP9 and has lower ligand efficiency of − 0.31 kcal/mol for MMP2 and − 0.28 kcal/mol for MMP9 (Fig. 6a, b). These values are found to be higher than the positive control, batimastat. Since batimastat has binding free energy (∆G) of − 6.34 kcal/mol for MMP2 and − 5.66 kcal/mol for MMP9. These binding energy values indicate the leading conformational position of the different phytoligand molecules and (∆G) was derived from the sum of torsional and intermolecular free energy in AutoDock. The negative value of (∆G) signifies robust favorable bonding between phytoligands and the MMPs. From results, it implies that the phytoligand, indigocarpan captures its good number of flattering conformation towards MMPs. Various interactions occur between the indigocarpan and MMPs. Active residues such as GLY418, ALA419, ILE424, ARG101, GLU404, TYR425, THR426, THR428, fits into the binding pocket of MMP2 to mediate van der Waals interactions (Fig. 6a, b). PRO423, LEU399, CYS102, and PRO417 were found to possess hydrogen bonding interactions. VAL400, LEU508, LEU191, LEU420, and PHE431exhibits pi-alkyl interactions and HIS403 possesses pi-pi stacked interactions (Fig. 7a, b). Similarly, in MMP9, TYR50, TYR54, ASP182, PHE181, LEU187, and VAL101, residues of MMP9 interact with indigocarpan through van der Waals interactions, (Fig. 6c, d) and conventional hydrogen bonding interactions using residues such as ARG51, and PRO180. Pi-Alkyl interactions occur by residues viz. HIS190 and PRO102 (Fig. 7c, d).

Fig. 6.

Batimastat binds with active regions of MMP2 which was represented by ribbon (a). Enlarged view of MMP2-batimastat binding complex (b). Indigocarpan binds with active regions of MMP2 which was represented by ribbon (c). Enlarged view of MMP2-Indigocarpan binding complex (d). Batimastat binds with active regions of MMP9 which was represented by ribbon (e). Enlarged view of MMP9-batimastat binding complex (f). Indigocarpan binds with active regions of MMP9 which was represented by ribbon (g). Enlarged view of MMP9-Indigocarpan binding complex (h)

Fig. 7.

Molecular docking simulation using AutoDock 4.2. a 2D diagram shows the interaction of indigocarpan with active residues of MMP2. b Similarly, 2D diagram shows the interaction of batimastat (positive control) with amino acid residues of MMP2. Molecular docking simulation using AutoDock 4.2. c 2D diagram shows the interaction of indigocarpan with active residues of MMP9. d Similarly, 2D diagram shows the interaction of batimastat (positive control) with amino acid residues of MMP9. Molecular docking simulation using AutoDock Vina.  e 2D diagram shows the interaction of indigocarpan with active residues of MMP2. f Similarly, 2D diagram shows the interaction of batimastat (positive control) with amino acid residues of MMP2. 7 Molecular docking simulation using AutoDock Vina. g 2D diagram shows the interaction of indigocarpan with active residues of MMP9. h Similarly, 2D diagram shows the interaction of batimastat (positive control) with amino acid residues of MMP9

In MMP2, ALA422 interacts with benzofuran ring through alkyl bonding and HIS403, LEU191, and VAL400 interacts with chromene moiety through alkyl bonding of the indigocarpan molecule and inhibits its activity. Similarly, in MMP9, HIS190 and TYR179 interact with the benzofuran ring and ARG51, PRO180 and LEU187 interacts with chromene moiety of the indigocarpan molecule. This interaction of indigocarpan mediates strong binding to the active pocket of both MMP2 and MMP9 and inhibits its functions. HIS403 and ALA422 were found to be the major active residues in MMP2 and HIS190, TYR179, PRO180, and ARG51 were found to the major active residues in MMP9 for their inhibition. The presence of methoxy functional group at C3 and C9 position have been recognized as a significant structural prerequisite for the cytotoxic property of the indigocarpan molecule compared with other phytoligands. Predominantly, the presence of methoxy group in the benzofuran ring recognizes the phytoligand (indigocarpan) to fit within the major active residues of matrix metalloproteinases to inhibit its functions.

The MMP inhibitory effects of various phytocompounds have been investigated using AutoDock vina software. Using this program, the docking of major phytoligands was done with the active pockets of both MMP2 and MMP9. Presence of Zn²⁺ ions enhances the activity of both MMPs by activating the catalytic domain. Here, the indigocarpan compound acts as a non-zinc binding MMP inhibitor that binds with the active regions of both MMPs, but it does not interact with the Zn²⁺ ions, which indicates that the indigocarpan compound possesses the specificity and selectivity to bind deeply within the active pocket and induces a specific protein conformation to inhibit MMPs activity. Docking studies provide various configurations that enable to examine favorable binding modes. Among the five compounds investigated, indigocarpan, erythroxdiol and mucronulatol possess greater affinity towards MMP2 and MMP9 and the compounds exhibit higher potential when compared with batimastat, positive control (Fig. 6e, f). From results, it has been found that indigocarpan has a binding affinity of − 7.68 kcal/mol towards MMP2 and -6.35 kcal/mol towards MMP9, whereas batimastat (positive control) possesses a binding affinity of − 6.34 kcal/mol towards MMP2 and − 5.66 kcal/mol towards MMP9 (Tables 4, 5) (Fig. 6g, h). In MMP2, TYR182 interacts with benzofuran ring through alkyl bonding and PRO105 interacts with chromene moiety through alkyl bonding of the indigocarpan molecule and inhibits its activity. Similarly, in MMP9, VAL637 and ARG634 interact with the benzofuran ring and PRO97 interacts with chromene moiety of the indigocarpan molecule (Fig. 7e, f). This interaction of indigocarpan mediates strong binding to the active pocket of both MMP2 and MMP9 and inhibits its functions. TYR182 and PRO105 were found to be the major active residues in MMP2 and VAL637, ARG634 and PRO97 were found to the major active residues in MMP9 for their inhibition (Fig. 7g, h).

Table 4.

Comparative results of molecular docking using AutoDock 4.2 and AutoDock Vina – Binding affinity of MMP2 with phytoligands

S. no.	Compounds	Molecular Docking—AutoDock 4.2
		Binding energy (kcal/mol)	Amino acid residues
1	Mucronulatol	− 7.40	LYS99,PRO100,TYR395,ASP392,GLY394, TYR427,THR426,TYR425,THR511,ILE424, LEU507
2	Indigocarpan	− 7.68	LEU191,CYS102,LEU399,PRO423,ALA419, GLY418,ALA192,THR428,PHE431,GLU404, VAL400,HIS403,TYR425,ARG101,LEU508
3	Indigocarpan diacetate	− 5.13	ILE424,THR511,TYR427,THR426,
4	Erythroxydiol X	− 7.07	LYS99,THR511,GLU393,ASP392,PRO100, TYR426,TYR425,TYR395,ILE424,TYR42
5	Erythroxydiol Y	− 7.13	TYR427,THR511,LYS99,THR426,ILE424, TYR425,TYR395,GLY394,ASP392
6	Batimastat (Positive control)	− 7.20	GLN393,TYR427,SER546,PHE512,ARG495, THR426,ILE424,ASP392,LYS372,PRO100, TYR425,LYS99,ARG98

S. no.	Compounds	Molecular Docking—AutoDock Vina
		Binding energy (kcal/mol)	Amino acid residues
1	Mucronulatol	− 5.59	ARG146,TYR232,LEU215,VAL220, PHE235,LYS234,GLN219
2	Indigocarpan	− 8.00	THR56,ASN111,PRO105,GLY103,HIS193, PRO183,LEU190,THR182,PHE184
3	Indigocarpan diacetate	− 7.50	ARG146,PHE431,LEU433,ARG43
4	Erythroxydiol X	− 7.70	VAL150,LEU433
5	Erythroxydiol Y	− 7.60	PHE431
6	Batimastat (Positive control)	− 7.20	PRO100,ASP392,LYS99,PRO514, ALA510,TYR427,GLN393

Table 5.

Comparative results of molecular docking using AutoDock 4.2 & AutoDock Vina – Binding affinity of MMP9 with phytoligands

S. no.	Compounds	Molecular Docking—AutoDock 4.2
		Binding energy (kcal/mol)	Amino acid residues
1	Mucronulatol	− 6.08	ASP182,PRO182,TYR50,ARG51,TYR179, TYR54,HIS190,LEU187,GLY100,VAL101
2	Indigocarpan	− 6.35	TYR50,HIS190,PRO102,TYR54,TYR179, GLN108,VAL101,GLY100,LEU187,PRO180, PHE181,ASP182,ARG51
3	Indigocarpan diacetate	− 4.94	TYR54,TYR179,GLN108,VAL101,GLY100, LEU187,PRO180,PHE181,ASP182,HIS190, ARG51
4	Erythroxydiol X	− 5.85	TYR50,PRO180,ASP182,LEU187,ARG51, PRO102,GLY100,TYR52
5	Erythroxydiol Y	− 6.20	TYR54,PRO102,VAL101,TYR179,HIS190, LEU187,ARG51,TYR50
6	Batimastat (Positive control)	− 5.66	TYR50,ARG51,GLU47,ASN38,PRO180, PHE181,ALA173,ASP182,ARG36,GLY183, LYS184,ASP34,LEU35

S.no.	Compounds	Molecular Docking—AutoDock Vina
		Binding energy (kcal/mol)	Amino acid residues
1	Mucronulatol	− 7.80	GLN640,ASP643,VAL637,ARG634,VAL589, TYR590
2	Indigocarpan	− 8.50	VAL587,TRP633,VAL589,PRO97,TYR590, ARG634,VAL637,VAL642,GLN640
3	Indigocarpan diacetate	− 8.20	ARG645,ARG634,PHE635,PRO97,THR96, ASP185,ILE556
4	Erythroxydiol X	− 8.30	VAL587
5	Erythroxydiol Y	− 7.60	GLY545,SER548,LEU409
6	Batimastat (Positive control)	− 7.60	PRO97,ARG95,ARG634,VAL637,LEU555, TRP633,TYR590,VAL587

From these results, it is clear that indigocarpan acts as an MMP inhibitor than batimastat (positive control) concerning its binding efficiency and inhibitory potential against matrix metalloproteinases to abate metastasis which is confirmed by both open source softwares and online tools (Selvaraj et al. 2016; Ajeet et al. 2018).

Toxicity prediction

Evaluation of ADME and toxicity properties of the drug molecules is an essential factor in the drug discovery process. Here, all the phytoligands have been analyzed to determine their in silico pharmacokinetic parameters using SwissADME to evaluate their drug-likeness, physicochemical properties, lipophilicity, water solubility, and its potency. In silico pharmacokinetics exhibit GI absorption, blood–brain barrier (BBB) penetration, P-gp substrate, and cytochrome P450 for the phytoligands, and the respective values have been shown in Table 6. One among the PK factors is cytochrome P450, which plays a key role in drug metabolism and comprises a heme-containing protein family mediates several xenobiotic substances, drug molecules, and carcinogenic factors. CYP 1, 2, and 3 families were mainly involved in the biotransformation of drugs and chemical substances. Different isoforms of CYPs have been reported such as CYP2C19, CYP2C9, CYP2D6, and CYP3A4. In particular, the CYP3A4 enzyme is a main considerable factor that is concerned with the xenobiotic metabolism that occurs in the human system. This enzyme is found to be the highest level in the human liver (~ 40%) which catalyzes about 50% of the clinical drug molecules. From the results, the physicochemical properties, lipophilicity, water-solubility, PK properties, and drug-likeness were examined for the phytoligands and batimastat (positive control) (Table 7). Indigocarpan, mucronulatol, and erythroxydiol X compounds possess drug-likeness and lead likeness potentiality as they obey Lipinski rule. Log P_o/w (iLogP) is the n-octanol/ water partition coefficient is a main component of the physicochemical parameter in drug discovery (Daina et al. 2014). Here, the Log P value falls < 5 for all the phytoligands, which is found to be an acceptable value for the drug molecules for penetration into the membranes (Table 8). Indigocarpan, mucronulatol, and indigocarpan diacetate were found to be highly water-soluble, erythroxdiol X and erythroxydiol Y are moderately soluble and batimastat (positive control) is poorly soluble in water. GI absorption, BBB penetration, and P-gp were found to be normal for all the phytoligands, but batimastat has less GI absorption and lack of BBB-permeability (Table 9). Among the cytochrome P450, CYP3A4 metabolizes the drug to the target site and the drug executes its mechanism of action (Roy et al. 2009). Similarly, admetSAR results have shown the toxicity profiling of the phytoligands and batimastat. All the phytoligands do not have a carcinogenic effect, and compounds such as indigocarpan, mucronulatol, and erythroxydiol X do not have Ames mutagenicity (Ujan et al. 2019) (Table 10). Thus, from SwissADME and admetSAR results, compound indigocarpan possesses fine ADME properties and does not show any toxicity effects compared with batimastat.

Table 6.

SwissADME – physiological properties

S. no.	Compounds	Mol.wt (g/mol)	No of rotatable bonds	No of H-bond acceptors	No of H-bond donors (Å)	TPSA
1	Mucronulatol	302.32	3	5	2	68.15
2	Indigocarpan	316.31	2	6	2	77.38
3	Indigocarpan diacetate	400.38	6	8	0	89.52
4	Erythroxydiol X	280.36	2	2	1	37.30
5	Erythroxydiol Y	306.48	2	2	2	40.46
6	Batimastat (Positive control)	477.64	15	4	4	161.07

Table 7.

SwissADME – Solubility properties

S. no.	Compounds	Log Po/w (iLOGP)	Consensus Log Po/w	Log S (ESOL)	Class
1	Mucronulatol	2.69	2.58	−3.75	Soluble
2	Indigocarpan	2.52	2.09	−3.46	Soluble
3	Indigocarpan diacetate	3.62	2.79	−3.77	Soluble
4	Erythroxydiol X	2.59	2.97	−2.92	Soluble
5	Erythroxydiol Y	3.47	4.19	−4.64	Soluble
6	Batimastat (positive control)	2.32	3.07	−4.30	Soluble

Table 8.

SwissADME – Pharmacokinetics

S. no.	Compounds	GI-Absorption	BBB Permability	P-gp	CYP1A2 Inhibitor	CYP2C19 Inhibitor	CYP2C9 Inhibitor	CYP2D6 Inhibitor	CYP3A4 Inhibitor
1	Mucronulatol	High	Yes	Yes	Yes	No	No	Yes	Yes
2	Indigocarpan	High	Yes	Yes	Yes	No	No	Yes	Yes
3	Indigocarpan diacetate	High	No	No	Yes	Yes	Yes	Yes	Yes
4	Erythroxydiol X	High	Yes	No	No	No	Yes	No	No
5	Erythroxydiol Y	High	Yes	No	No	No	Yes	No	No
6	Batimastat (positive control)	High	No	Yes	No	Yes	Yes	No	Yes

Table 9.

SwissADME – Druglikeness properties

S. no.	Compounds	Lipinski rule	Violoation	Bioavailability score	Leadlikeness
1	Mucronulatol	Yes	0	0.55	Yes
2	Indigocarpan	Yes	0	0.55	Yes
3	Indigocarpan diacetate	Yes	0	0.55	No
4	Erythroxydiol X	Yes	0	0.55	Yes
5	Erythroxydiol Y	Yes	0	0.55	No
6	Batimastat (positive control)	Yes	0	0.55	No

Table 10.

admetSAR – Toxicity predictions

S. no.	Compounds	Carcinogenicity	hERG inhibition	Ames Mutagenicity	Acute oral toxicity (Kg/mol)
1	Mucronulatol	–	–	–	1.937
2	Indigocarpan	–	–	–	1.774
3	Indigocarpan diacetate	–	–	+	1.899
4	Erythroxydiol X	–	–	–	2.599
5	Erythroxydiol Y	–	–	–	1.794
6	Batimastat (positive control)	–	+	–	2.635

Conclusion

From this study, it has been revealed that indigocarpan inhibits the activity of the matrix metalloproteinases, MMP2, and MMP9. By intending MMP2 and MMP9 as drug targets, which have a vital role in ECM degradation to regulate metastasis, indigocarpan could act as a potent anti-metastatic drug candidate by inhibiting MMP2 and MMP9 to abate cancer metastasis. However, further in vitro experiments have to address the effects of indigocarpan with matrix metalloproteinases.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

NDK conceived and designed the research. NDK and PS conducted the experiments and analyzed the data. PS wrote the manuscript. Both authors read and approved the manuscript.

Declarations

Conflict of interest

The authors declare that they have no conflict of interest in the publication.