ADME prediction and toxicity:
We previously synthesized an existing library of phenylhydrazones with potent antimalarial activity against chloroquine-sensitive and chloroquine-resistant P. falciparum (3D7 and Dd2) strains [33]. However, because of the presence of imine and nitro groups, the most potent compounds, PHN3 and PHN6, showed unfavourable toxicity predictions. Here, by simply swapping the nitro groups on PHN3 and PHN6 with other electron-withdrawing groups (CN, CHO, COCH3, COOH, CONH2 and SO3H) and reducing the imine group, we designed 72 analogues, as presented in Fig. S1. Since the majority of compounds fail to reach clinical trials due to their undesirable pharmacokinetic profiles [34] and poor drug-like and safety properties pose significant obstacles in the drug development process [36], we evaluated all the designed analogues of PHN3 and PHN6 in silico to determine their physicochemical, pharmacokinetic and safety profiles.
This phase was critical because it made it possible to choose analogues that are optimal, nontoxic, and have good pharmacokinetic and physicochemical characteristics. In accordance with SwissADME and admetsar, 12 designed analogues demonstrated favourable physicochemical, pharmacokinetic, and drug-like characteristics, suggesting that they could be promising drug candidates for further investigation. We observed that all analogues containing sulfonic acid and aldehyde groups were eliminated due to their violation of the Brenk rule, making them inappropriate electron withdrawing groups to substitute for the nitro groups in the phenylhydrazone scaffold. Analogues bearing various combinations of CN, COCH3, COOH, and CONH2 groups exhibited good physicochemical and pharmacokinetic properties and could be suitable substitutes for nitro groups in phenylhydrazone scaffolds. However, few of them were predicted to inhibit various forms of cytochrome P450 enzymes according to SwissADME and the Admetsar server and were eliminated because their inhibition of these enzymes could induce or inhibit the metabolism of other drugs, resulting in clinically significant drug–drug interactions [Table S4, Table S5 and Table S6]. Drugs are actively transported across biological membranes by the membrane protein p-glycoprotein (P-gp) [37]. P-gp can facilitate or restrict the absorption, distribution, excretion, and toxicity of many medications [38]. Several pharmacokinetic drug‒drug interactions can also occur when two compounds are combined, one of which is a substrate and the other is an inhibitor of the transporter [39]. The propensity of the optimal analogues to produce drug‒drug interactions or be effluxed from the cell is limited because they are not p-gp substrates or inhibitors [40]. The compounds were also very soluble, had good absorption in the human intestine and exhibited high gastrointestinal absorption. They did not cross the blood brain barrier. Furthermore, each of these drugs conforms to the Lipinski, Ghose, Verber, Egan and Muegge rules for orally active drugs and hence has the potential to be developed into oral medications. Apart from hepatotoxicity, drug-induced cardiotoxicity is another often-reported adverse event that has resulted in drug withdrawal [41]; one source of drug-induced cardiotoxicity is inhibition of hERG (human ether-à-go-go-related gene) K+ channels, which produces a type of fatal arrhythmia known as torsade de pointes or long QT syndrome [42]. Analysis of the optimal analogues suggested that they are neither cardiotoxic nor hepatotoxic (Table S4 and Table S5). The best analogues were further used for molecular docking and dynamics studies for protein‒ligand characterization.
Validation of the docking protocol:
Initial verification of the docking methodology was carried out using pyrimethamine (inhibitor) and its binding protein (PDB ID 3QGT) prior to investigating the binding modes and energies of the designed compounds with PfATP6 and its mutant protein. The inhibitor was removed from P. falciparum dihydrofolate reductase (PDB ID 3QGT) and redocked into the active site using autodock tools. Using PyMOL 2.3 software, these poses were superimposed on top of the reference inhibitor, as shown in Fig S2. The estimated root mean square deviation (RMSD), which was 0.03, indicates little variation. The docked conformation was also evaluated using Discovery Studio Visualizer, which verified the presence of comparable active site residues that facilitate hydrophobic contact and hydrogen bonding between the inhibitor and binding protein. A docking approach with a greater degree of reliability was ensured by the computed root mean square deviation (RMSD) in conjunction with similar interactions since a value of less than 2 angstroms indicates good reproducibility of the cocrystallized structure [43].
Molecular docking studies
The receptor protein PfATP6, also known as PfSERCA, has been demonstrated to be a common target of artemisinin-based antimalarials [44]. A total of 1228 amino acids constitute this 139 kDa protein. Since the protein's three-dimensional structure is unavailable, we predicted the PfATP6 three-dimensional structure using the Swiss model server [45]. After evaluating the built model with PROCHEK (https://saves.mbi.ucla.edu/), the malarial PfATP6 mutant type L263E was then modelled. Fig S3 Summary of the Ramachandran plots of the studied proteins.
In the docking procedure, the grid coordinates were generated by enclosing the residues at positions 263, 264, 267, 977, 981, 985, 1039, 1040, 1041 and 1042 for the malarial PfATP6 proteins and at positions ASP54, CYS15, ILE14, LEU164, ASN108, PHE58, PRO113, ILE112 and MET55 for the PfDHFR proteins in a receptor grid box. These residues have been investigated and confirmed as the active site residues mediating the binding of artemisinin to PfATP6 and pyrimethamine to PfDHFR [46, 47]. Additionally, the chemical structures of artemisinin and pyrimethamine were adequately prepared and docked into the binding pockets of PfATP6 and PfDHFR as control drugs, as shown in Fig. 3 and Fig. 4.
Based on the molecular interactions between the control drugs and wild-type proteins, it can be inferred that the main mechanism of inhibition is van der Waals interactions. For instance, conventional hydrogen bonds with LEU1040 and ASN1039 and hydrophobic contacts with LYS260, PHE264, ILE1041 and LEU1046 stand out as fundamental amino acids involved in the inhibition of wild-type PfATP6 by artemisinin. In addition, pyrimethamine has been shown to form hydrogen bonds with the amino acids ILE14, CYS15, ASP54, ILE164, and THR185 and hydrophobic interactions with the amino acids ALA16, MET55, PHE58 and ILE112 for the catalytic inhibition of wild-type PfDHFR. However, analysis of the interactions of artemisinin and pyrimethamine with the PfATP6 and PfDHFR mutant proteins revealed fewer hydrogen bonds and fewer hydrophobic interactions. For example, no conventional hydrogen bonds with active site residues were observed after the analysis of all the conformations of artemisinin with the PfATP6 mutant protein. Furthermore, pyrimethamine formed three hydrogen bonds with active site residues during the analysis of its best docked pose with the qm-PfDHFR complex, while the wild-type PfDHFR and the same drug exhibited six conventional hydrogen bonds with active site residues. In addition, the molecular docking simulation revealed that the interaction of artemisinin and pyrimethamine with the mutant proteins resulted in greater free binding energy and inhibition constant (Ki) values than those of the wild-type proteins. These findings suggest that the propensity of Plasmodium falciparum to switch amino acid residues at these positions may modify the binding mechanism, perhaps reducing the susceptibility of the organism to artemisinin and pyrimethamine.
Table S1 shows the estimated free energy of binding for the compounds and control drugs against the targets, which presented the best scoring of the compounds. According to the obtained results, the designed compounds present suitable affinity towards the four proteins, with compounds B24 and B36 having the best performance against the studied proteins.
When the designed compounds were docked with wild-type and mutant proteins, all compounds hit the four targets at the active site mainly via van der Waals interactions. The interactions of the control drugs with the targets were compared with those of the designed compounds to obtain further insight into the role of molecular interactions in the ligand‒receptor complex. Interestingly, the compounds engaged with both wild-type PfATP6 and PfDHFR and their mutants via a greater number of conventional hydrogen bonds as well as some notable hydrophobic interactions with the active site residues, as indicated in Figs. 4 and 5. Additionally, when bound to both targets, B24 and B36 exhibited lower inhibition constants (Ki) (below 5 µM) for all the studied proteins, indicating that it is possible that they will inhibit these targets more successfully than the control drugs.
Molecular dynamics study
To assess the stability of the malarial protein‒ligand complexes formed from the docking runs, molecular dynamics (MD) simulations were carried out using the gromacs 2023 program on a Linux system [48]. The steps included the preparation of protein and ligand topologies with the pdbgmx tool and CHARMM, respectively; solvation; the addition of ions; energy minimization; NVT and NPT equilibration; and production and analysis. The optimal docked poses of B24 and B36 were taken into consideration because these ligands displayed the best van der Waal interaction with the target active site residues, had lower binding free energies, and had lower Ki values than the control drugs. We first examined the stability of artemisinin and pyrimethamine upon binding to their target proteins and compared the results with those of the proposed compounds (B24 and B36). The root mean square deviation (RMSD) of the wild-type and mutant malarial PfATP6 and DHFR proteins with respect to the control drugs are shown in Fig. S4 and Fig. S5. The average backbone RMSD of artemisinin bound to malarial PfATP6 and its mutant protein throughout the whole simulation period was 0.54126 and 0.630925 nm, respectively; thus, a greater fluctuation was observed for the L263E-artemisinin complex than for the wild-type malarial PfATP6, as indicated in Fig. S3. The RMSD for the backbone was calculated for pyrimethamine in complex with the wild-type and quadruple mutant proteins over a 50 ns simulation. The average RMSD of pyrimethamine-qmDHFR was slightly lower (0.141858 nm) than that of pyrimethamine-wild-type protein (0.206869 [Table S1]). However, the RMDS for the entire 50 ns simulation remained below 0.4 nm, as illustrated in Fig. S3 and Table S1.
Hydrogen bond analysis revealed that fewer hydrogen bonds formed between the control drugs and mutant proteins during the simulation period (Figs. S6 and S7). This was also evident for the malarial PfATP6 mutant treated with artemisinin, in which there were few and sporadic hydrogen bonds (S7 Fig). This indicates that there is possibly poor complementarity between artemisinin and this target protein. Figure 7 shows the backbone RMSD analysis of B24 and B36 bound to the dihydrofolate reductase quadruple mutant enzyme and the PfATP6 mutant, respectively. The average RMSD of qmDHFR was 0.151612 and 0.195304 for B24 and B36, respectively, and 0.660084 and 0.356839 nm for the malarial PfATP6 mutant protein. This suggests that B24 and B36 protein complexes are stable and less likely to induce structural instability with these proteins.
Analysis of the number of hydrogen bonds formed during the entire simulation revealed that B24 and B36 formed a maximum of eight and five hydrogen bonds with the mutant PfATP6 protein and a maximum of six and seven hydrogen bonds with the quadruple mutant DHFR. Additionally, the formation of hydrogen bonds between B24 and B36 and the mutant proteins was frequent and prominent. The ability of the two compounds to maintain a stable number of hydrogen bonds during the simulation period could suggest a strong binding affinity towards the mutant protein compared to the control drugs. The results of the gmx hbond analysis of the two compounds and the mutant proteins are shown in Fig. 8.