Molecular docking-based virtual screening, molecular dynamic simulation, and 3-D QSAR modeling of some pyrazolopyrimidine analogs as potent anti-filarial agents

Abstract

Lymphatic filariasis and onchocerciasis are common filarial diseases caused by filarial worms, which co-habit symbiotically with the Wolbachia organism. One good treatment method seeks Wolbachia as a drug target. Here, a computer-aided molecular docking screening and 3-D QSAR modeling were conducted on a series of Fifty-two (52) pyrazolopyrimidine derivatives against four Wolbachia receptors, including a pharmacokinetics study and Molecular Dynamic (MD) investigation, to find a more potent anti-filarial drug. The DFT approach (B3LYP with 6-31G** option) was used for the structural optimization. Five ligand-protein interaction pairs with the highest binding affinities were identified in the order; 23_7ESX (-10.2 kcal/mol) > 14_6EEZ (− 9.0) > 29_3F4R (− 8.0) > 26_6W9O (− 7.7) ≈ doxycycline_7ESX (− 7.7), with good pharmacological interaction profiles. The built 3-D QSAR model satisfied the requirement of a good model with R² = 0.9425, Q²_LOO = 0.5019, SDEC = 0.1446, and F test = 98.282. The selected molecules (14, 23, 26, and 29) perfectly obeyed Lipinski’s RO5 for oral bio-availability, and showed excellent ADMET properties, except 14 with positive AMES toxicity. The result of the MD simulation showed the great stability associated with the binding of 23 onto 7ESX’s binding pocket with an estimated binding free energy (MM/GBSA) of − 60.6552 kcal/mol. Therefore, 23 could be recommended as a potential anti-filarial drug molecule, and/or template for the design of more prominent inhibitors.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40203-022-00136-y.

Introduction

Lymphatic Filariasis (LF) also known as elephantiasis and Onchocerciasis (river blindness) are common Neglected Tropical Diseases (NTD), which are caused by some parasitic nematode worms (Sightsavers 2013). LF is caused by filarial worms like Wuchereria bancrofti, Brugia timori and Brugia malayi, which are been transmitted by mosquitoes, while Onchocerca volvulus is the causative agent for onchocerciasis, which is transmitted from one person to another by blood-feeding black flies (Bakowski et al. 2019). Elephantiasis alone is responsible for not less than 2.8 million disabilities globally (Jacobs et al. 2019). The global program intended to eliminate these filarial diseases started far back through the Mass Drug Administration (MDA) of anti-filarial such as ivermectin, albendazole, and diethylcarbamazine, either as a dual (annual to bi-annual) or as triple-drug (once every 3 years) treatment (Jacobs et al. 2019; Carter et al. 2020). However, it became unlikely that the MDA regimen will be adequate to eliminate these filarial diseases in all endemic areas, majorly due to their inability to kill the macrofilariae (Lakshmi et al. 2010). Given the current scenario, therefore, a macrofilaricidal agent is required to kill worms to reduce both diseases’ elimination time frames (Sashidhara et al. 2014).

Fortunately, one unique characteristic of these filarial worms is their symbiotic co-existence with a known bacterium referred to as Wolbachia (Slatko et al. 2010). In the search for new anti-filarial drugs, some researchers have chosen the option of targeting Wolbachia, which past research has shown that its elimination from the host filarial nematodes leads to antifilarial effects with the reduction of adult worm’s lifespan (Bouchery et al. 2013; McGillan 2017). Although the anti-bacteria drug, doxycycline has been used clinically for the treatment of filarial diseases over the years, the treatment method is not efficient enough for use in mass administration including requirements for long treatment periods (4–6 weeks) as well as contraindications in pregnancy and children (McGillan 2017). Therefore, advances in the development of new anti-Wolbachia agents with short treatment periods and reduced complications are necessary.

Some compounds of the pyrazolopyrimidine class were earlier reported to show a variety of bioactivities such as anti-viral agents, anti-malarial, anti-depressants, anti-tuberculosis, and kinase inhibitors (McGillan et al. 2021; Ugbe et al. 2022a). However, certain side effects have been associated with most of the drugs in this class such as hypnotic and/or anxiolytic effects. To further explore the anti-filarial effect of the pyrazolopyrimidine compounds, McGillan (2017) synthesized several pyrazolopyrimidine derivatives and reported their inhibitory activities against Wolbachia infected insect cells (Aedes albopictus, C6/36). Notable targets of Wolbachia pipientis include Oxidoreductase α-DsbA1 (PDB ID: 3F4R), OTU deubiquitinase (6W9O), thiol-disulfide exchange protein alpha-DsbA2 (6EEZ), and Cytoplasmic incompatibility factor CidA (7ESX) amongst others.

Computer-aided drug design plays a crucial role in the discovery of new drug molecules in pharmaceutical design, drug metabolism, and medicinal chemistry. It saves time, and cost and tends to be highly effective for the evaluation of a large virtual database of chemical compounds (Adeniji et al. 2020). Molecular docking simulation computer-aided screening method which probes the binding of ligands in the active sites of the protein target using a valid docking tool (Ibrahim et al. 2020). Pharmacokinetics analysis on the other hand is important in the pre-clinical study of new drug compounds to ascertain how such drug compounds affect the living organism when administered. Some of the most important pharmacokinetic properties to be determined during pre-clinical testing include Absorption, Distribution, Metabolism, Excretion, and Toxicity (ADMET) (Lawal et al. 2021; Ibrahim et al. 2021). Physico-chemical properties such as molecular weight, Topological Polar Surface Area (TPSA), lipophilicity, hydrogen bond donors, and hydrogen bond acceptors amongst others are necessary to predict a drug’s likelihood of being orally bioavailable (Lipinski et al. 2001). This work focuses on the virtual molecular docking screening of a series of Fifty-two (52) pyrazolopyrimidine derivatives against Four (4) Wolbachia targets, 3-D QSAR modeling, Molecular Dynamics (MD) simulation, and prediction of pharmacokinetic properties of some selected analogs, to find a more potent drug molecule which would be suitable for the treatment of filarial diseases.

Materials and methods

Data acquisition

A series of Fifty-two (52) pyrazolopyrimidine derivatives with reported bioactivities (EC₅₀ in nM) against Wolbachia-infected insect cells (Aedes albopictus, C6/36), were sourced from the literature (McGillan 2017). The various bioactivity (EC₅₀) values were separately converted to pEC₅₀ using Eq. (1) (Ugbe et al. 2022a). The molecular structures of the various pyrazolopyrimidine derivatives were shown in Online Resource 1.

$${\mathit{pEC}}_{50}={-log}_{10}\left({\mathit{EC}}_{50},\times ,{10}^{-9}\right)$$ 1

Ligand preparation

The molecular structures of all the compounds were drawn using the ChemDraw Ultra, saved as MDL molfile format, and thereafter imported separately onto the Spartan ’14 Graphical User Interface while enabling the auto conversion of 2-D models to 3-D. The imported molecules were initially subjected to energy minimization and then saved in Spartan file format. The resulting structures were then fully optimized first by using Molecular Mechanics Force Field (MMFF) and thereafter Density Functional Theory (DFT) with Becke’s three-parameter read-Yang-Parr hybrid (B3LYP) option and utilizing the 6-31G basis set. The optimized structures were then saved as PDB and SD file formats for subsequent use in molecular docking and 3-D QSAR studies respectively (Wang et al. 2020; Ugbe et al. 2021).

Preparation of the protein receptors

The crystal structures of Four (4) Wolbachia target proteins (PDB codes: 3F4R, 6EEZ, 6W9O, and 7ESX) were retrieved from the RCSB Protein Data Bank in PDB file format, and then prepared separately using the Molegro virtual docker by eliminating water molecules, cofactors and co-crystallized ligands contained within the protein structures (Ugbe et al. 2022b). The various receptors used in the virtual docking screening were described in Table 1.

Table 1.

Description of enzymes used for the docking screening

S. No	Enzyme	Organism	PDB ID	Resolution (Å)
1	Alpha-DsbA1	Wolbachia	3F4R	1.60
2	Alpha-DsbA2	Wolbachia	6EEZ	2.25
3	OTU deubiquitinase	Wolbachia	6W9O	1.47
4	Cytoplasmic incompatibility factor CidA	Wolbachia	7ESX	1.80

PDB ID – 3F4R, 6EEZ, 6W9O, 7ESX

Molecular docking-based screening

Molecular docking investigation was performed separately between the Four (4) different receptors of Wolbachia pipientis and all 52 compounds, including the reference drug (Doxycycline) using the Auto Dock Vina of PyRx v software tool (Ugbe et al. 2021). The screening was conducted to ascertain the most active pyrazolopyrimidine compounds against the various protein targets. PyRx calculates the binding affinities of the receptor-ligand interactions which are necessary to describe how fit the molecules bind to the target protein. A more negative binding affinity will indicate a greater chance of the potential drug molecule to initiate protein biochemical action/reaction (Kumar et al. 2016).

Evaluation of pharmacokinetic properties

Predicting pharmacokinetics properties plays a critical role in the early stage of drug discovery. This is because only molecules which demonstrate good ADMET and drug-likeness properties reach the pre-clinical research phase (Ugbe et al. 2021). Therefore, Four (4) pyrazolopyrimidine analogs (14, 23, 26, and 29) having the highest binding scores with 6EEZ, 7ESX, 6W9O, and 3F4R respectively were subjected to drug-likeness and ADMET tests using two online web servers; http://www.swissadme.ch/index.php and http://biosig.unimelb.edu.au/pkcsm respectively. Lipinski’s rule of five (RO5) also called the Pfizer rule is a well-established provision for determining the oral bioavailability of a given compound (Lipinski et al. 2001; Lawal et al. 2021). Consequently, these analogs were subjected to the RO5 criterion to ascertain their oral bioavailability.

Molecular dynamics simulation and MM/GBSA calculation

Molecular dynamics (MD) simulation of 7ESX_23 complex was performed using the combined approach of Chemistry at Harvard Macromolecular Mechanics (CHARMM) force field, Nano-scale Molecular Dynamics (NAMD), and Visual Molecular Dynamics (VMD). The CHARMM-GUI, an established web-based platform that utilizes the CHARMM force field, was used to generate the input files for the simulation by NAMD (Lee et al. 2016). The periodic boundary condition was utilized while fitting the system into a cubic water box for solvation. The protein was solvated and neutralized explicitly in an aqueous solution of 0.10 M concentration of potassium chloride salt (Edache et al. 2022). To stabilize the complex structure and to ensure steric clashes will not result, energy minimization was performed. The resulting system of ions and solvent was then equilibrated to stabilize the system at a temperature chosen for the simulation (310.15 K) at a constant number of particles, volume, and temperature (NVT ensemble), and to stabilize the pressure by keeping the number of particles, pressure, and temperature (NPT ensemble) constant using 100ps time frame (Muniba 2019). MD was then performed on the resulting system for 1ns (500,000 steps), while the results were visualized using VMD and the Biovia discovery studio, all on an HP laptop computer; Processor (Intel(R) Core(TM) i5-4210U CPU @ 1.70 GHz 2.40 GHz), Installed RAM (8.00 GB), System type (64-bit operating system, x64-based processor), Edition (Windows 10 Home Single Language), Version 21H2. A similar procedure was described elsewhere (Edache et al. 2022). Additionally, MolAICal software was used to compute the ligand-binding affinity by Molecular Mechanics Generalized Born Surface Area (MM/GBSA) method based on the resulting MD log files obtained with NAMD (Bai et al. 2020). MM/GBSA is estimated using Eqs. (2)–(4) (Bai et al. 2020).

$${\Delta G}_{\mathit{bind}}=\Delta H-T\Delta S\approx {\Delta E}_{\mathit{MM}}+{\Delta G}_{\mathit{sol}}-T\Delta S$$ 2

$${\Delta E}_{\mathit{MM}}={\Delta E}_{\mathit{internal}}+{\Delta E}_{\mathit{ele}}+{\Delta E}_{\mathit{vdw}}$$ 3

$${\Delta G}_{\mathit{sol}}={\Delta G}_{\mathit{SA}}+{\Delta G}_{\mathit{GB}}$$ 4

Where, ∆E_MM and −T∆S represent respectively the gas phase MM energy and conformational entropy. ∆E_MM contains electrostatic ∆E_ele, van der Waals energy ∆E_vdw and ∆E_internal of bond, angle, and dihedral energies. ∆G_sol is the solvation free energy equal to the sum of the nonelectrostatic solvation component ∆G_SA and electrostatic solvation energy ∆G_GB.

3 – D QSAR modeling

The alignment of molecular structures plays a critical role in 3D-QSAR modeling (Al-Attraqchi and Mordi 2022) as it strongly determines the predictive accuracy and statistical quality of any given 3D-QSAR model (ElMchichi et al. 2020). Different alignment methods have been reported previously such as atom-based, docking-based, pharmacophore-based, and co-crystallized conformer-based alignments amongst others (Zhang et al. 2020; Al-Attraqchi and Mordi 2022). In this study, the atom-based alignment was adopted using the Open3DAlign (O3A) tool. The atom-based method attempts to match the atoms of the various structures to be aligned with those of the template structure, based on the atom’s properties such as the partial charge.

The aligned structures were used for building the 3-D QSAR model using the Open3DQSAR software (Zhang et al. 2020). The Comparative Molecular Field Analysis (CoMFA) which is concerned with steric and electrostatic fields’ contributions was studied (ElMchichi et al. 2020). A dataset of 52 compounds was divided into a training set and a test set of 36 and 16 molecules respectively, i.e. percentage ratio of 70:30. The steric and electrostatic Molecular Interaction Fields (MIFs) analysis was carried out on the aligned compounds placed within a 3-D cubic lattice of grid size 1.5 Å and a 5.0 Å out gap (Tosco and Balle 2011). Variables pretreatment was carried out as follows; energy cut-off (30.0 kJ/mol), elimination of variables having constant or near-constant values, and standard deviation cut-off (level = 2.0) (Al-Attraqchi and Mordi 2022). The Un-informative Variable Elimination-Partial Least Square (UVE-PLS) was used to build the statistical model and for generating the steric and electrostatic contour plots (Edache et al. 2022). The resulting model was then cross-validated using the Leave-One-Out (LOO), Leave-Two-Out (LTO), and Leave-Many-Out (LMO). The steric and electrostatic contour maps were visualized on Maestro v. 12.3.

Results and discussion

Virtual docking screening

The results (binding affinities) of the docking simulation conducted between the Four (4) receptors of Wolbachia pipientis and the various pyrazolopyrimidine derivatives, as well as the reference drug (Doxycycline), were reported in Table 2.

Table 2.

Summary of binding affinities of interactions between pyrazolopyrimidine derivatives and different Wolbachia pipientis receptors used for the target fishing

Comp ID	Protein-ligand binding affinities (kcal/mol)
	3F4R	6EEZ	6W9O	7ESX
1	− 7.0	− 8.5	− 7.1	− 8.3
2	− 6.9	− 8.1	−  6.8	− 8.0
3	− 7.4	− 8.3	− 7.2	− 9.4
4	− 7.6	− 8.3	− 7.3	− 8.1
5	− 7.2	− 8.8	− 7.3	− 8.0
6	− 7.1	− 8.6	− 7.1	− 7.8
7	− 6.9	− 8.1	− 7.0	− 8.2
8	− 6.9	− 8.4	− 7.2	− 8.5
9	− 7.2	− 7.9	− 7.5	− 8.2
10	− 7.0	− 7.7	− 7.0	− 8.0
11	− 6.9	− 8.0	− 7.0	− 8.0
12	− 7.0	− 8.4	− 7.3	− 7.9
13	− 7.1	− 8.0	− 7.4	− 7.5
14	− 7.5	− 9.0	− 7.3	− 8.8
15	− 7.4	− 8.6	− 6.8	− 8.1
16	− 7.3	− 8.7	− 6.9	− 8.7
17	− 7.2	− 8.3	− 7.4	− 8.6
18	− 6.9	− 8.4	− 7.5	− 8.4
19	− 6.8	− 7.5	− 6.9	− 8.7
20	− 7.4	− 7.6	− 7.3	− 8.5
21	− 7.0	− 7.6	− 7.1	− 8.4
22	− 7.3	− 8.6	− 6.9	− 7.8
23	− 7.4	− 7.4	− 7.0	− 10.2
24	− 7.1	− 6.9	− 7.2	− 8.0
25	− 6.6	− 7.4	− 7.3	− 7.3
26	− 7.7	− 8.2	− 7.7	− 8.1
27	− 7.6	− 8.2	− 7.3	− 8.9
28	− 7.8	− 8.5	− 7.6	− 9.0
29	− 8.0	− 8.1	− 7.7	− 7.8
30	− 7.3	− 8.0	− 7.2	− 7.8
31	− 7.1	− 7.5	− 7.2	− 7.8
32	− 6.3	− 6.2	− 5.8	− 7.7
33	− 6.3	− 6.6	− 6.0	− 7.6
34	− 6.3	− 6.3	− 6.3	− 7.2
35	− 6.7	− 6.6	− 6.5	− 7.5
36	− 6.9	− 6.6	− 6.8	− 7.4
37	− 5.9	− 6.8	− 5.8	− 7.2
38	− 6.4	− 6.4	− 6.6	− 7.4
39	− 6.7	− 7.3	− 6.5	− 7.9
40	− 6.7	− 7.6	− 6.6	− 8.9
41	− 7.2	− 7.7	− 6.8	− 7.7
42	− 6.6	− 7.9	− 7.0	− 8.2
43	− 6.8	− 7.6	− 6.9	− 8.2
44	− 6.8	− 7.5	− 7.3	− 7.8
45	− 6.2	− 6.3	− 6.0	− 8.0
46	− 7.1	− 8.5	− 7.3	− 7.5
47	− 6.7	− 7.0	− 7.0	− 7.7
48	− 6.8	− 7.6	− 6.5	− 7.9
49	− 6.8	− 7.4	− 6.8	− 8.1
50	− 6.9	− 7.2	− 6.9	− 7.4
51	− 6.8	− 7.1	− 6.7	− 8.7
52	− 6.8	− 6.5	− 6.1	− 7.5
Ref	− 6.9	− 6.9	− 7.4	− 7.7

PDB ID – 3F4R, 6EEZ, 6W9O, 7ESX, Ref reference drug (Doxycycline)

. It can be observed from Table 2 that no particular ligand best interacted with all the studied receptors combined. That is, a ligand may bind very strongly with a given receptor but shows a weak interaction with another receptor. However, Four (4) ligand-protein interaction pairs with the greatest negative binding scores were identified in the order; compound 23 with 7ESX (-10.2 kcal/mol)> 14 with 6EEZ (− 9.0 kcal/mol)> 29 with 3F4R (− 8.0 kcal/mol) > 26 with 7ESX (− 7.7 kcal/mol). Also, no ligand-protein interaction pair involving the reference drug (Doxycycline) was identified that could compare with the identified interaction pairs, except doxycycline_7ESX complex with a binding score of − 7.7 kcal/mol equal to that of 26_7ESX complex. Therefore, the virtual screening was effective and subsequent discussion shall be based on these more active molecules (Table 3).

Table 3.

Molecular structures of some selected pyrazolopyrimidine analogs

Comp ID	Molecular structures
14
23
26
29
Doxycycline

The pharmacological interactions between the receptors’ amino acid residues and the selected compounds (14, 23, 26, and 29) as well as the reference drug (Doxycycline) were summarized in Table 4, while the 2D and 3D views of the binding interactions as adapted from the Discovery Studio Visualizer were shown in Figs. 1, 2, 3, 4, and 5. This was to provide insight into the mode of binding of these ligands with the active sites of the various target proteins.

Table 4.

Predicted binding interaction profiles of 14, 23, 26, 29, and Doxycycline with the receptors

Complex	Binding affinity (kcal/mol)	Amino acid	Bond type	Interaction	Distance (Å)
7ESX_23	− 10.20	GLU-188	Hydrogen bond	Conventional hydrogen bond	2.01, 3.05
		LYS-232	Hydrogen bond	Conventional hydrogen bond	2.68, 2.91
		ASN-77	Hydrogen bond	π-donor hydrogen bond	2.96
		GLU-191	Electrostatic	π-anion	4.06
		PHE-228	Hydrophobic	π- π T shaped	5.22
		ARG-74	Hydrophobic	π-sigma	3.57
		TRP-37	Hydrophobic	π-alkyl	5.39
		LEU-75	Hydrophobic	π-alkyl	5.44
		ARG-74	Hydrophobic	π-alkyl	4.30, 5.49
		ARG-36	Hydrophobic	Alkyl	4.95
		LEU-75	Hydrophobic	Alkyl	4.64
6EEZ_14	− 9.00	LYS-155	Hydrogen bond	Conventional hydrogen bond	2.30, 2.57
		TYR-89	Hydrogen bond	Carbon hydrogen bond	2.99
		LYS-118	Hydrogen bond	Carbon hydrogen bond	3.49
		PHE-159	Hydrophobic	π- π stacked	3.85
		TYR-89	Hydrophobic	π- alkyl	5.47
3F4R_29	− 8.00	ASP-103	Hydrogen bond	Conventional hydrogen bond	2.94
		LYS-109	Hydrogen bond	Conventional hydrogen bond	2.41, 2.67
		ASP-103	Hydrogen bond	Carbon hydrogen bond	3.73
		ASN-106	Hydrophobic	π-sigma	3.97
		ALA-110	Hydrophobic	π-alkyl	4.63
		ALA-110	Hydrophobic	Alkyl	3.64
6W9O_26	− 7.70	GLU-81	Hydrogen bond	Carbon hydrogen bond	3.62
		ARG-131	Electrostatic	π-cation	3.61
		GLU-135	Electrostatic	π-anion	4.44
		PHE-82	Hydrophobic	π- π T shaped	4.71
		PRO-88	Hydrophobic	π-sigma	3.64
		TRP-90	Hydrophobic	π-alkyl	4.88
		PRO-88	Hydrophobic	π-alkyl	5.43
		PRO-88	Hydrophobic	Alkyl	4.25
		LYS-85	Hydrophobic	Alkyl	4.74
		LYS-85	Donor-donor	Unfavorable	2.01
7ESX_	− 7.70	ILE-288	Hydrogen bond	Conventional hydrogen bond	2.03
Doxycycline		TYR-251	Hydrogen bond	Conventional hydrogen bond	2.56
		LEU-243	Hydrogen bond	Conventional hydrogen bond	2.67
		LYS-246	Hydrogen bond	Conventional hydrogen bond	2.09
		LYS-248	Hydrogen bond	Conventional hydrogen bond	2.03
		PHE-289	Hydrogen bond	Conventional hydrogen bond	2.81
		SER-290	Hydrogen bond	Carbon hydrogen bond	3.51
		LYS-287	Hydrogen bond	Carbon hydrogen bond	3.22
		SER-244	Hydrogen bond	Carbon hydrogen bond	3.03
		LYS-287	Hydrophobic	π-alkyl	5.50
		VAL-250	Donor-donor	Unfavorable	1.17

ALA alanine, ARG arginine, ASN asparagine, ASP aspartic acid, GLU glutamic acid, ILE isoleucine, LEU leucine, LYS lysine, PHE phenylalanine, PRO proline, SER serine, TRP tryptophan, TYR tyrosine, VAL valine

Fig. 1.

Binding interaction between 23 and Cytoplasmic incompatibility factor CidA (PDB: 7ESX)

Fig. 2.

Binding interaction between 14 and Alpha-DsbA2 (PDB: 6EEZ)

Fig. 3.

Binding interaction between 29 and Alpha-DsbA1 (PDB: 3F4R)

Fig. 4.

Binding interaction between 26 and OTU deubiquitinase (PDB: 6W9O)

Fig. 5.

Binding interaction between doxycycline and Cytoplasmic incompatibility factor CidA (PDB: 7ESX)

These compounds were said to interact very adequately with the respective target receptors as shown by the presence of hydrogen bonding (H-bond), hydrophobic interactions, and in some cases electrostatic interactions. (Table 4). However, more interactions were visible from the binding profile of compound 23 with 7ESX, involving a total of Four (4) conventional H-bonds, One (1) π-donor H-bond, One (1) π-anion electrostatic interaction, and up to Eight (8) hydrophobic interactions. Four groups can be identified in the molecular structure of compounds 23 as pyridine, pyrimidine, pyrazole, and benzoate groups, all interacting significantly with the receptor’s amino acid residues. The carbonyl group (C = O) oxygen of the benzoate group formed 2 H-bonds with LYS-232 at interaction distances of 2.68 and 2.91 Å. The remaining 2 conventional H-bonds were formed by GLU-188 with the pyridine group and the linker amine group at 2.01 Å and 3.05 Å respectively. Also, the π-donor H-bond was between ASN-77 and the pyrazole π-system at 2.96 Å. Visible were the π-anion interactions between the π-electrons systems of GLU-191 and the benzoate group at 4.06Å. Several hydrophobic interactions were formed including π- π T shaped with PHE-228 (5.22 Å), π-sigma with ARG-74 (3.57 Å), π-alkyl with TRP-37 at 5.39 Å, LEU-75 at 5.44 Å, and ARG-74 at 4.30 Å and 5.49Å, and alkyl interactions with ARG-36 and LEU-75 at distances of 4.95 Å and 4.64 Å respectively. It is important to note that no unfavorable interaction was seen in the 23_7ESX binding interaction profile (Fig. 1). The complex involving the reference drug, doxycycline_7ESX on the other hand showed more H-bonding interactions than 23_7ESX, consisting of a total of Six (6) conventional H-bonds and Three (3) Carbon-H-bonds. Only One (1) hydrophobic interaction was however visible. More so, an unfavorable donor-donor clash with VAL-250 was formed (Fig. 5). Therefore, compound 23 exhibited stronger and safer binding interactions with the Cytoplasmic incompatibility factor CidA than the reference drug (doxycycline)

Evaluation of pharmacokinetic properties

Drug-likeness analysis and ADMET study were conducted on the Four (4) compounds (14, 23, 26, and 29) to ascertain their oral bioavailability. The results of both investigations were presented in Tables 5 and 6 respectively, while Fig. 6 shows their Boiled Egg’s representation.

Table 5.

Predicted drug-likeness properties of some selected pyrazolopyrimidine derivatives

Comp ID	MW (g/mol)	TPSA (Å2)	MLOGP	Log S (ESOL)	HBD	HBA	RO5	PAINS	BRENK	SA
14	359.40	55.11	3.56	− 4.60	1	4	0	0	0	3.28
23	387.43	81.41	2.79	− 4.49	1	5	0	0	0	3.26
26	437.54	72.93	3.26	− 5.47	1	4	0	0	0	3.67
29	442.51	84.65	2.23	− 4.28	1	5	0	0	0	3.58

MW molecular weight, TPSA topological polar surface area, ESOL estimated solubility, HBD hydrogen bond donors, HBA hydrogen bond acceptors, RO5 Lipinski rule of five violation, SA synthetic accessibility score

Table 6.

Predicted ADMET properties of some selected pyrazolopyrimidine derivatives

Comp ID	Absorption		Distribution		Metabolism				Excretion	Toxicity
	HIA (%)	Skin LogKp	BBB LogBB	CNS LogPS	CYP34A		CYP2D6		Total clearance	AMES	MRTD
					S	I	S	I
14	95.55	− 2.74	0.325	− 2.060	YES	YES	NO	NO	0.246	YES	− 0.253
23	98.01	− 2.74	− 0.801	− 2.439	YES	YES	NO	NO	0.701	NO	− 0.087
26	98.62	− 2.73	− 0.804	− 2.022	YES	YES	NO	NO	0.791	NO	0.172
29	96.12	− 2.75	− 0.146	− 2.953	YES	YES	NO	NO	0.591	NO	0.147

BBB Blood brain barrier, CNS Central nervous system, HIA Human intestinal absorption, Skin skin permeability, LogBB the logarithmic ratio of brain to plasma drug concentration, LogPS blood-brain permeability-surface area product, CYP34A/CYP2D6 cytochrome p450 isoforms, S substrate, I inhibitor, MRTD Maximum recommended tolerated dose

Fig. 6.

The boiled-egg representation of compounds 14, 23, 26, and 29

Lipinski’s RO5 for oral-bioavailability has been widely applied in the discovery of new drug molecules (Ugbe et al. 2022b). It asserts that a drug molecule may likely not be orally bio-available when it has Hydrogen Bond Donors (HBD) of greater than 5, Hydrogen Bond Acceptors (HBA) > 10, Molecular Weight (MW) > 500, and lipophilicity (MLOGP > 4.15 or WLOGP > 5) (Lipinski et al. 2001). Whenever a molecule passed at least three of the four provisions of the RO5, it is said to comply with Lipinski’s rule for oral bioavailability (Lawal et al. 2021). Table 5 showed that all the tested pyrazolopyrimidine derivatives passed the drug-likeness test (Lipinski RO5) by showing no violation. The reported values of Topological Polar Surface Area (TPSA) for the molecules were less than 140 Ų. Also, the values of the synthetic accessibility (SA) scores of these compounds were less than 5.00 (easy portion on a scale of 1–10), suggesting easy laboratory synthesis. The predicted values of the estimated water solubility (Log S) are in the range of − 4 > Log S > − 6, indicating these molecules are moderately soluble. The compounds were equally estimated to be free from pains and brenk alerts.

The estimated ADMET properties reported in Table 6, showed a very high Human Intestinal Absorption (HIA) (greater than 90%) for all tested compounds. Skin permeability is a key factor in transdermal drug delivery development. Values of skin permeation constant LogKp > − 2.50 indicates poor skin permeability. As a result, the various compounds tested showed LogKp values < − 2.50, connoting good skin permeability. Drug molecule penetration through the Blood-Brain Barrier (BBB) and Central Nervous System (CNS) comes with certain criteria. To enable a drug molecule penetrates the BBB and CNS readily, the logarithmic ratio of brain to plasma drug concentration (logBB) must be > 0.3 and the blood-brain permeability-surface area product (logPS) be > − 2 respectively. Consequently, only 14 with logBB of 0.325 readily penetrate the BBB as also indicated by its location within the boiled egg’s yolk shown in Fig. 6, while the various compounds are non-CNS permeable. Also, 23, 26, and 29 were located in the Boiled Egg’s white, an indication that they were predicted to be passively absorbed by the gastrointestinal tract.

Furthermore, some group of enzymes called cytochrome P450 enzymes are important in the body to facilitate drug metabolism and to help in their excretion. The two major isoforms enhancing drug metabolism, CYP-34 A and CYP-2D6 were tested. The tested molecules are not substrates and inhibitors of CYP2D6 but are both substrates and inhibitors of CYP3A4, an indication of a well-moderated metabolic process. Figure 6 showed that only compound 23 was predicted not to be effluated from the central nervous system by P-glycoprotein. P-glycoprotein acts as a biological barrier by extruding toxins and xenobiotics, including drugs out of cells. The extent of drug removal from the body is determined by the drug’s total clearance. The range of values of total clearance for all the tested molecules is good. Additionally, all the compounds except 14 showed no AMES toxicity, implying that they are non-mutagenic and cannot act as carcinogens. Also available in Table 5 is the Maximum Recommended Tolerated Dose (MRTD) predicted for the various molecules. MRTD value of ≤ 0.477 log (mg/kg/day) is considered low, while a value > 0.477 log (mg/kg/day) is considered high. The overall drug-likeness and ADMET properties of the selected compounds showed good pharmacokinetic profiles, except compound 14 which showed positive AMES toxicity. Therefore, these molecules could be considered potential drug candidates for the treatment of filarial diseases.

Molecular dynamics simulation

To analyze the dynamics of the protein-ligand interaction, MD simulation was performed on the best protein-ligand interaction pair (23_7ESX complex) for 1ns (1000 ps) of chemical time (500,000 iterations). The results of this simulation as plots of Root-Mean-Square Deviation (RMSD), Root-Mean-Square Fluctuation (RMSF), Solvent Accessible Surface Area (SASA), and Radius of gyration (Rg) versus the time in ps were presented in Figs. 7, 8 and 9, and 10 respectively.

Fig. 7.

The plot of RMSD versus time for MD simulation of 23 with 7ESX

Fig. 8.

The plot of RMSF versus time for MD simulation of 23 with 7ESX

Fig. 9.

The plot of SASA versus time for MD simulation of 23 with 7ESX

Fig. 10.

The plot of radius of gyration versus time for MD simulation of 23 with 7ESX

The average RMSD value was estimated as 1.6801 Å which showed that the protein-ligand complex deviated only a little from its original conformation during the trajectory. The deviation was maximum during the first 100ps of the simulation, after which it drops and tends to remain slightly unstable until a further drastic drop in the RMSD at 1000 ps, an indication that the system was fast attaining stability and nearing equilibrium (Edache et al. 2020). RMSF is more like a calculation of the flexibility or the extent of movement of individual residue during a simulation. As seen from Fig. 8, the RMSF tends to drop as the simulation nears 1000ps, a further indication that the system was fast attaining stability. The SASA is simply the surface area that is in contact with the solvent in which the complex resides. From Fig. 9, it can be observed that the SASA only fluctuates slightly between 10.50 Ų and 11.6 Ų during the trajectory, an indication of stability (Edache et al. 2022). The Rg is the measure of the degree of compactness of a protein during the trajectory. Decreasing Rg indicates reducing residues’ flexibilities and more stability for the protein. Throughout the trajectory, the Rg varies between 27.283 Å and 28.365 Å which is equivalent to a difference of approximately 1.0 Å for the complex studied, connoting slight changes in the protein compactness as the simulation progresses, and therefore means the stability of the complex. Furthermore, it will not be complete without inspecting the simulated complex for possible protein-ligand interactions. As a result, the simulated complex was visualized using the Biovia discovery studio and the resulting binding interaction of 23 with the active site of 7ESX is presented in Fig. 11.

Fig. 11.

Binding interaction between 23 and Cytoplasmic incompatibility factor CidA (PDB: 7ESX) after MD simulation

The binding interaction pattern of the simulated complex (Fig. 11) deviated significantly from that of the non-simulated complex (Fig. 1) as several interactions majorly the hydrophobic interactions, electrostatic, and π-donor H-bond were lost. However, a significant number of important interactions were visible including Two (2) conventional H-bonding with SER-187 and ASN-77 at interaction distances of 2.32 Å and 1.94 Å respectively, Two (2) carbon H-bonding with ASN-77 and LEU-75 at 2.96 Å and 2.75 Å respectively. Others are hydrophobic interactions with ARG-36 and ARG-74 at 4.12 Å and 4.66 Å respectively. Additionally, no unfavorable steric bumps or clashes were visible. Furthermore, the result of binding free energy (MM/GBSA) computed for 23_7ESX by MolAICal is shown in Table 7.

Table 7.

Binding free energy parameters of 23_7ESX complex

Parameter	Value (kcal/mol)
∆E(internal)	+ 15.1432
∆E(electrostatic) + ∆G(solvation)	− 25.7373
∆E (Van der Waal)	− 50.0611
∆G binding (MM/GBSA)	− 60.6552 ± 0.528

The negative value of the estimated binding free energy (MM/GBSA) of the complex (− 60.6552 kcal/mol) indicates the favorability of the ligand-protein binding. Also, Vander Waals energy (− 50.0611 kcal/mol) contributed most to the binding free energy of the complex, connoting that Vander Waal/hydrophobic interactions played a crucial role in the binding process (Xu et al. 2019). It can therefore be inferred that compound 23 binds readily with the Cytoplasmic incompatibility factor CidA even within a dynamically perturbed system, and hence could be considered as a potential drug candidate for the treatment of filariasis.

3 – D QSAR modeling

Molecular structural alignment represents a key factor in ascertaining the predictive strength of a built 3-D QSAR model. Figure 12 (a–b) shows the molecular structure of the alignment template (compound 30) and the aligned structures as obtained from the super-imposition of the remaining 51 molecules on the template. The UVEPLS approach was used to develop the model. Some significant statistical parameters calculated for the model were presented in Table 8. Reported in Table 9 were the experimental pEC₅₀, predicted pEC₅₀, and their residuals together with their O3A scores. Additionally, a plot showing the correlation between predicted and experimental activities for both training and test sets was obtained and presented in Fig. 13. Also, the CoMFA model equation was summarized graphically as 3D contour maps as shown in Figs. 14 (a–b) and 15 (a–b).

Fig. 12.

Molecular alignment of structures for the QSAR modeling a Alignment template (compound 30 with the highest O3A_Score of 9057.78); b All structures aligned

Table 8.

Statistical parameters of the built model

Parameters		(UVEPLS)
PC		5
R2		0.9425
SDEC		0.1446
F test		98.282
Q2LOO		0.5019
SDEPLOO		0.4253
Golbraikh and Tropsha acceptable model criteria
r2	0.7501	r2 > 0.60
$\left|r_o^2,-,r_o^{\prime 2}\right|$	0.1074	$\left|r_o^2,-,r_o^{\prime 2}\right|$ < 0.3
$\left|{(r}^2,-,r_0^2,)/,r^2\right|$	0.00123	$\left|{(r}^2,-,r_0^2,)/,r^2\right|$ < 0.1
k	1.0457	0.85 < k < 1.15
Field contributions
Steric		0.5093 (50.93%)
Electrostatic		0.4907 (49.07%)

PC principal components, SDEP standard error of prediction, F test Fischer’s statistics, LOO leave one out, Q² cross-validated correlation coefficient, R² Correlation coefficient, SDEC standard error of a correlation, k slope of the plot of predicted activity against experimental activity, r² square correlation coefficients of the plot of experimental activity versus predicted activity values, r_o² square correlation coefficients of the plot of experimental activity versus predicted activity values at zero intercept, r′_o² square correlation coefficients of the plot of predicted activity versus experimental activity at zero intercept

Table 9.

Observed pEC₅₀, predicted pEC₅₀, residuals, and Open3DAlign scores of the various pyrazolopyrimidine derivatives

Comp ID	EC50 (nM)	pEC50	Pred. pEC50	Residuals	O3A Score
1	647	6.189	5.934	0.255	8591.09
2	1854	5.732	5.946	– 0.214	8602.81
3	3012	5.521	5.562	– 0.041	8306.61
4*	5176	5.286	6.022	– 0.736	8231.71
5	1384	5.859	5.865	– 0.006	8554.70
6*	145	6.839	6.146	0.693	8567.27
7*	90	7.046	6.338	0.708	7866.65
8*	93	7.032	5.936	1.096	8524.64
9*	33	7.481	6.426	1.055	8346.95
10	183	6.737	6.782	– 0.045	8389.24
11	518	6.286	6.286	0.00	8865.76
12	456	6.341	6.282	0.059	8561.81
13	2500	5.602	5.481	0.121	8035.49
14	93	7.032	6.782	0.25	8992.31
15	51	7.292	6.986	0.306	8755.17
16	15	7.824	7.98	– 0.156	8571.70
17	674	6.171	6.35	– 0.179	8757.04
18	664	6.178	6.178	0.00	8998.60
19	143	6.845	6.747	0.098	8895.54
20	664	6.178	6.382	– 0.204	8888.54
21	84	7.076	7.139	– 0.063	8551.79
22	102	6.991	7.082	– 0.091	8161.13
23*	179	6.747	6.346	0.401	8923.84
24	175	6.757	6.764	– 0.007	8943.94
25*	131	6.883	6.762	0.121	8918.56
26	1479	5.83	5.808	0.022	9053.70
27*	119	6.925	6.538	0.387	8743.78
28	43	7.367	7.351	0.016	7901.16
29	680	6.167	6.233	– 0.066	9022.08
30	844	6.074	6.055	0.019	9057.78
31*	1228	5.911	5.961	– 0.05	8789.76
32	105	6.979	7.041	– 0.062	6769.31
33	176	6.754	6.823	– 0.069	6961.39
34*	122	6.914	6.025	0.889	6443.42
35	164	6.785	6.731	0.054	7026.31
36	251	6.6	6.602	– 0.002	6816.61
37*	629	6.201	6.148	0.053	6770.61
38	311	6.507	6.449	0.058	7194.85
39	52	7.284	7.259	0.025	7723.79
40	73	7.137	7.186	– 0.049	7960.74
41	561	6.251	6.414	– 0.163	8114.80
42*	38	7.42	6.499	0.921	7833.08
43	16	7.796	7.752	0.044	8375.87
44	1183	5.927	5.892	0.035	7708.13
45	201	6.697	6.831	– 0.134	6377.11
46	38	7.42	7.313	0.107	7915.01
47	293	6.533	6.85	– 0.317	8661.85
48	32	7.495	7.094	0.401	7986.54
49*	14	7.854	7.465	0.389	7198.67
50*	23	7.638	7.013	0.625	6738.02
51*	34	7.468	7.465	0.003	7198.68
52*	49	7.31	7.206	0.104	7380.57

*Test set compounds

Fig. 13.

Correlation between predicted and experimental pEC₅₀ for training and test sets

Fig. 14.

Steric field contour maps of compound 23 a Blue contours represent regions of favorable steric bulk; b Red contours showing regions of unfavorable steric bulk

Fig. 15.

Electrostatic field contour maps of compound 23 a Green contours showing regions of unfavorable high electron density or favorable to electron-withdrawing groups; b Yellow contours represent regions favored by high electron density or unfavorable to electron-withdrawing substituents

The alignment process involves an early step that provided all the 52 compounds the opportunity of being chosen as the alignment template based on the compound with the highest Open3DAlign (O3A) score. The O3A scores of the various compounds were included in Table 8. Compound 30 had the highest O3A score of 9057.78 and hence was selected as the template upon which the remaining structures were superimposed. The model’s statistical parameters were computed for Five (5) Principal Components (PC) amongst which the fifth PC (PC = 5) performed relatively better with R² value of 0.9425, SDEC = 0.1446, and F-test = 98.282. The statistical parameters available in Table 9 were those associated with PC 5. The predictive strength of the regression models on new datasets of compounds can be estimated by cross-validation (Grohmann and Schindler 2008). A cross-validated coefficient of correlation (Q²) ≥ 0.50 indicates a good QSAR model. Here, three (3) types of Q² were calculated; Leave-one-out (LOO), Leave-two-out (LTO), and Leave-many-out (LMO), together with their associated Standard Error of Prediction (SDEP). Only Q²_LOO (0.5019) passed this criterion and was reported alone.

A linear correlation between the CoMFA descriptors (independent variables) and the activity values (dependent variables) was established by the PLS analysis method. The lower residual values between the predicted and observed activity values (Table 8) shows a strong predictive strength of the model. This was supported by the clustering of points along the lines of best fit in the plots of predicted pEC₅₀ versus the experimental pEC₅₀ (Fig. 13). This observation was supported by the conformation of the model to the Golbraikh and Tropsha criteria (Table 5) (Roy et al. 2016). The CoMFA QSAR equation is summarized graphically as a 3D contour map, which shows the regions within the molecules’ 3-D structural space where steric and electrostatic fields are associated with extreme values. The underlying principle behind CoMFA is that variations in the shape and strength of non-covalent interaction fields surrounding the molecules, such as steric or electrostatic fields can be related to changes in binding affinities (Kakarla et al. 2016). Therefore, molecular fields are key factors in binding affinity. The steric and electrostatic field contributions were 50.93% and 49.07% respectively (Table 9).

From the steric field contour maps available in Fig. 14 (a–b), the red contours represent regions of unfavorable steric bulk, while the blue contours show regions of favorable steric bulk. Regions in which steric bulk may reduce activity or affinity of the compound include positions 3 and 4 on the pyridine group, position 5 on the pyrimidine group, and position 2 on the benzoate group (Fig. 14b). For example, substituting the methyl group on position 5 of the pyrimidine group with a more bulky group like ethyl, isopropyl or tert-butyl could reduce the activity or binding affinity of the compound. On the other hand, more steric bulk favorable regions were identified (Fig. 14a), which include position 6 in the pyrimidine group, position 6 in the pyridine group, and position 2 in the benzoate group. This implies that the introduction of bulky substituent groups at these positions will improve the inhibitory activity of the molecule. From the electrostatic field contour maps available in Fig. 15(a–b), yellow contours represent regions favored by high electron density or unfavorable to electron-withdrawing substituents, while the green contours represent regions of unfavorable high electron density or favorable to electron-withdrawing groups. Five (5) regions in which the introduction of electron-withdrawing groups could reduce the inhibitory activity or binding affinity include all positions in the pyridine group, positions 5 and 6 in the pyrimidine group, position 2 in the pyrazole group, and the carbonyl group of the benzoate moiety (Fig. 15b). Also, regions of unfavorable high electron density were visible around the benzene ring system of the benzoate group and between the linker amine group and the pyrazole hetero atom. These regions need not be too electron-dense, hence electron-withdrawing groups will keep these regions at a low electron density which in turn will enhance the molecule’s inhibitory activity or binding affinity. In general, contour map analysis serves as a guide to designing new molecules with improved potency by adhering to the information encoded in the contour maps.

Conclusion

In this study, a molecular docking-based virtual screening, pharmacokinetics analysis, molecular dynamic simulation, and 3-D QSAR modeling were performed on the pyrazolopyrimidine derivatives. The molecular docking screening was effective as the Five (5) best protein-ligand interaction pairs were identified and ranked as 23_7ESX (– 10.2 kcal/mol) > 14_6EEZ (– 9.0 kcal/mol) > 29_3F4R (– 8.0 kcal/mol) > 26_6W9O (– 7.7 kcal/mol) ≈ doxycycline_7ESX (– 7.7 kcal/mol). The selected analogs (14, 23, 26, and 29) all obeyed Lipinski’s RO5 for oral bio-availability and showed excellent ADMET properties except 14, with positive AMES toxicity. Results of the MD simulation showed the stability of the 23_7ESX complex, exhibiting a favorable ligand-protein binding process with an estimated ∆G binding (MM/GBSA) of – 60.6552 kcal/mol. The 3 – D QSAR (CoMFA) model was developed and found to satisfy the requirement for validation tests with R² value of 0.9425, Q²_LOO = 0.5019, SDEC = 0.1446, and F test = 98.282. The anti-Wolbachia activities of the various compounds were well predicted by the model. The analysis of the steric and electrostatic contour maps could provide a useful guide for the future design of more active analogs. Special emphasis on compound 23 because it appears to be consistent with the various employed validation protocols, being that it possessed the highest binding score, showed excellent pharmacokinetic properties, and binds pharmacologically well with the target protein (7ESX). Therefore, 23 could be considered as a potential filarial drug candidate, and/or template for the design of more prominent Wolbachia inhibitors.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ADMET

Absorption, distribution, metabolism, excretion, and toxicity

ALA

Alanine

ARG

Arginine

ASN

Asparagine

ASP

Aspartic acid

B3LYP

Becke’s three-parameter read-Yang-Parr hybrid

BBB

Blood brain barrier

CHARMM

Chemistry at Harvard macromolecular mechanics

CidA

Cytoplasmic incompatibility factor A

CNS

Central nervous system

CoMFA

Molecular field analysis

CPU

Central processing unit

CYP-34A/CYP-2D6

Cytochrome p450 isoforms

DFT

Density functional theory

EC₅₀

Half-maximal inhibitory concentration

ESOL

Estimated solubility

F test

Fischer’s statistics

GLU

Glutamic acid

HBA

Hydrogen bond acceptor

HBD

Hydrogen bond donor

HIA

Human intestinal absorption

HIS

Histidine

ILE

Isoleucine

LEU

Leucine

LMO

Leave many out

LogBB

Logarithmic ratio of brain to plasma drug concentration

LogPS

Blood-brain permeability-surface area product

LOO

Leave one out

LTO

Leave two out

LF

Lymphatic filariasis

LYS

Lysine

MD

Molecular dynamics

MDA

Mass drug administration

MIFs

Molecular interaction fields

MMFF

Molecular mechanics force field

MM/GBSA

Molecular mechanics generalized born surface area

MRTD

Maximum recommended tolerated dose

MW

Molecular weight

NAMD

Nano-scale molecular dynamics

NTD

Neglected tropical diseases

O3A

Open3D align

PC

Principal component

PDB

Protein data bank

pEC₅₀

Negative log of EC₅₀

PHE

Phenylalanine

PRO

Proline

QSAR

Quantitative structure activity relationship

Rg

Radius of gyration

RAM

Random access memory

RMSD

Root-mean-square deviation

RMSF

Root-mean-square fluctuation

RO5

Rule of five

SA

Synthetic accessibility

SDEC

Standard error of correlation

SDEP

Standard error of prediction

SEE

Standard error of estimation

SER

Serine

SASA

Solvent accessible surface area

TPSA

Topological polar surface area

TRP

Tryptophan

TYR

Tyrosine

UVE-PLS

Un-informative variable elimination-partial least square

VAL

Valine

VMD

Visual molecular dynamics

Author contributions

GAS and AU conceived and designed the study. FAU carried out the study and drafted the manuscript. IA conducted the technical editing. All authors read and approved the final manuscript.

Funding

No funding was received for this study.

Data availability

All data related to this study are included herein otherwise available on request.

Declarations

Conflict of interest

The authors declare that they have no competing interests.

Ethical approval and consent to participate

Not applicable.

Consent for publication

Not applicable.