An In Silico Study of the Interactions of Alkaloids from Cryptolepis sanguinolenta with Plasmodium falciparum Dihydrofolate Reductase and Dihydroorotate Dehydrogenase
The Plasmodium falciparum dihydrofolate reductase (PfDHFR) and dihydroorotate dehydrogenase (PfDHODH) are essential for Plasmodium falciparum growth and development, and have been validated as targets for the development of new antimalarial agents. Several alkaloids isolated from Cryptolepis sanguinolenta have been reported to have antiplasmodial activity, but their protein targets are unknown. Therefore, molecular docking and molecular dynamics simulations were used to investigate the interactions and stability of the alkaloids with PfDHFR and PfDHODH. Based on physicochemical characteristics, alkaloids were grouped as sterically bulky (sb) or planar (pg). Docking results revealed strong binding affinities (−6.0 to −13.4 kcal/mol) of the alkaloids against PfDHODH and various strains of PfDHFR while interacting with key residues such as Asp54 and Phe58 in PfDHFR. The pg alkaloids had high binding affinity and preference for the inhibitor binding domain over the flavin mononucleotide (FMN) binding domain in PfDHODH due to size considerations. From the molecular dynamics trajectories, protein-alkaloid complexes were stable throughout the simulation, with supporting evidence from root mean square deviations, root mean square fluctuations, radius of gyration, free binding energies, and other parameters. We report herein that biscryptolepine and cryptomisrine (sb class), as well as cryptolepinone, cryptoheptine, cryptolepine, and neocryptolepine (pg class), are capable of inhibiting PfDHFR effectively in pyrimethamine sensitive and resistant cells. Also, our results show that alkaloids of the pg class can inhibit PfDHODH as FMN decoys, as well as direct enzyme inhibitors, thereby halting crucial protein function.
Despite various efforts designed to eliminate malaria, the disease remains a formidable challenge to public health programs globally. According to the World Health Organization (WHO), an estimated 228 million cases of malaria were recorded globally in 2018. Of these cases, about 405,000 deaths were recorded. About 90 countries have been designated as malaria endemic regions, and thus about 40% of the world’s population are at risk of the disease. The WHO African region accounted for 93% of all cases recorded in 2018. Unfortunately, most malaria victims in sub-Saharan Africa are children under 5 years and pregnant women . The socioeconomic impact of the disease is enormous. Of the five Plasmodium species that cause malaria in humans, Plasmodium falciparum (P. falciparum) is responsible for most of the severe clinical malaria cases and is also the most lethal . Owing to the essential roles they play in various metabolic processes in the parasite and their minimal homology with host proteins, two proteins in P. falciparum have been validated as targets for malaria chemotherapy. These proteins are the P. falciparum dihydrofolate reductase (PfDHFR) and P. falciparum dihydroorotate dehydrogenase (PfDHODH) [3, 4].
P. falciparum dihydrofolate reductase (PfDHFR; Figure 1) is an essential enzyme in folate biosynthesis and is responsible for the generation of the DNA base, deoxythymidine monophosphate (dTMP). PfDHFR is also involved in the biosynthesis of purine nucleotides and amino acids (Hyde, 2005). Inhibition of PfDHFR by typical antifolates such as trimethoprim, cycloguanil, and pyrimethamine prevents biosynthesis of thymidine and consequently interrupts DNA biosynthesis, leading to parasite demise. Unfortunately, point mutations at certain amino acid residues such as Ala16, Ile51, Cys59, Ser108, and Ile164 in the active site of wild-type PfDHFR have led to widespread resistance of the P. falciparum parasite to these drugs [4, 5]. The discovery of new PfDHFR inhibitors to overcome drug-resistant parasites is therefore an urgent need. Parasitic protozoa possess an endogenous folate biosynthetic pathway that is susceptible to antifolate inhibitors and a salvage pathway to make use of exogenous folate. Humans, however, lack the ability to synthesize folates de novo and therefore rely on salvage pathways for their folate requirements. Protozoa possess dihydrofolate reductase-thymidylate synthase (DHFR-TS) as a bifunctional enzyme in which DHFR and TS are two domains of a single homodimeric protein; in humans, DHFR and TS occur as two separate, independently functional proteins. Human cells also depend on DHFR for DNA replication; therefore, developing inhibitors that are not only potent but also selective for the parasitic pathogen is of ultimate importance. Fortunately, the sequence of the DHFR gene has significantly diverged throughout evolution, generating unique protein sequences in diverse organisms .
Dihydroorotate dehydrogenase (DHODH; Figure 1) catalyzes the oxidation of dihydroorotate (DHO) to orotate by reducing flavin mononucleotide (FMN), and this is a key step in pyrimidine biosynthesis . Pyrimidines are essential metabolites in all cells and can be either synthesized or obtained from already formed pyrimidine bases via pyrimidine salvage enzymes. They are required not only for nucleic acid biosynthesis but also for the biosynthesis of phospholipids and glycoproteins. The de novo pyrimidine biosynthetic pathway is intact in most organisms, including Plasmodium. However, unlike mammalian cells, P. falciparum cannot salvage preformed pyrimidine bases or nucleosides and utilizes de novo pyrimidine biosynthesis exclusively to meet its metabolic requirements. The Plasmodium purine and pyrimidine metabolic pathways are distinct from those of their human hosts. In the absence of salvage enzymes for the incorporation of preformed bases and nucleosides, all of the de novo biosynthetic enzymes are presumably essential to Plasmodium species. The lack of sequence conservation in the inhibitor-binding site of DHODH across species suggests that the identification of species-selective inhibitors is feasible. Furthermore, while the lack of pyrimidine salvage pathways in Plasmodium makes it vulnerable to inhibition of this pathway, the presence of salvage pathways in human cells affords additional selectivity . The blockage of the P. falciparum dihydroorotate dehydrogenase (PfDHODH) enzyme shuts off the downstream production of biomolecules necessary for parasitic multiplication, leading to their death . Thus, targeting purine and pyrimidine metabolic pathways provides a promising route for novel malaria drug development .
Compounds such as quinine, chloroquine, mefloquine, and artemisinins have been the go-to drugs for malaria treatment. However, the protozoans have developed resistance against many of the current therapies and this has led to the exploration of various other sources for potential malaria therapeutic agents . In many indigenous African homes, plants have served as a mainstay for the treatment of malaria and malaria-like symptoms [11–15]. In fact, it is estimated that over 70% of the global population uses plant-based products for their healthcare needs . One such plant used in Ghana and many West African nations for malaria treatment is Cryptolepis sanguinolenta (C. sanguinolenta). C. sanguinolenta is a shrub indigenous to many countries in West and Central Africa used in the clinical therapy of malaria and urinary and upper respiratory tract infections. Numerous pharmacological activities have been demonstrated in extracts from the roots and leaves of C. sanguinolenta. Pharmacological activities demonstrated in in vitro and in vivo experiments include hypotensive, antipyretic, anti-inflammatory, antimalarial, antiviral, and antidiabetic effects. Alkaloids isolated from C. sanguinolenta have also been shown to possess antiplasmodial capabilities . The isolated alkaloids include cryptolepine, 11-isopropylcryptolepine, hydroxycryptolepine, neocryptolepine, cryptospirolepine, isocryptolepine, cryptolepicarboline, cryptomisrine, cryptolepinone, biscryptolepine, cryptoheptine, and cryptoquindoline. In in vitro antiplasmodial studies on the multidrug-resistant P. falciparum K1 strain, the dose of drug required to inhibit parasite growth by 50% (IC50) for most of these alkaloids was determined to be between 0.8 µM and 102 µM [18–20]. Cryptolepine has also been shown to exhibit promising synergistic interactions in vitro with amodiaquine (mean ΣFIC50 = 0.287 ± 0.10 µM), additive effects in combination with chloroquine and lumefantrine, and antagonistic effects with mefloquine . The mode of action of these alkaloids and the proteins they interact with, however, remain elusive. To develop promising therapeutics from these compounds, it is important to identify the targets that these compounds interact with. This information will aid in lead optimization efforts.
Computational approaches are widely used to provide insights into molecular events that mediate molecular recognition, enzyme catalysis, and ligand (agonist or antagonist) binding. Molecular docking and molecular dynamics (MD) studies, for instance, have been used widely in identification of lead compounds and understanding mechanisms of enzyme action and inhibition [22, 23]. Molecular docking has been used to identify potent inhibitors of Aurora kinase A (AKA), and this resulted in the identification of the (4-methoxy-pyrimidin-2-yl)-phenyl-amine scaffold as a promising warhead. Structure-based virtual screening, de novo design, and free energy perturbations also served as a starting point for the discovery of anticancer therapeutics . Molecular docking and MD simulations have also been used to tease out new targets for HIV-1 therapeutics . This work provides a plausible molecular basis for the antiplasmodial action of the alkaloids of C. sanguinolenta by exploring the interactions of the alkaloids with 2 validated drug targets of P. falciparum—PfDHFR and PfDHODH. Both wild-type and mutant proteins were used in the study. We herein report on the binding affinities, complex stabilities, and key molecular interactions of alkaloids from C. sanguinolenta with PfDHFR and PfDHODH. The results of the study suggest that biscryptolepine, cryptomisrine, cryptolepinone, cryptoheptine, cryptolepine, and neocryptolepine are capable of inhibiting PfDHFR effectively. The results also show that alkaloids such as cryptolepine, neocryptolepine, isocryptolepine, cryptoheptine, cryptolepinone, and hydroxycryptolepine likely inhibit PfDHODH by acting as flavin mononucleotide (FMN) decoys, and also as direct enzyme inhibitors, thereby halting crucial protein function.
2.1. Protein Target Selection and Preparation
The crystal structures of the protein targets were obtained from the Protein Data Bank (PDB, https://www.rcsb.org/). Three variants of PfDHFR were selected: wild-type PfDHFR (PDB ID: 3QGT, wt-PfDHFR hereafter) , a double mutant (PDB ID: 1J3J, Cys59Arg + Ser108Asn, dm-PfDHFR hereafter), and a quadruple mutant (PDB ID: 1J3K, Asn51Ile + Cys59Arg + Ser108Asn + Ile164Leu, qm-PfDHFR hereafter) . All three proteins were dimers, complexed with their cofactors, NADPH and dUMP. The wt-PfDHFR and dm-PfDHFR were both cocrystallized with pyrimethamine, whereas qm-PfDHFR was bound to a known inhibitor, WR99210 (6,6-dimethyl-1-[3-(2,4,5-trichlorophenoxy)propoxy]-1,6-dihydro-1,3,5-triazine-2,4-diamine). Each dimer had both a dihydrofolate reductase and a thymidylate synthase (TS) domain. Chain A was prepared for ligand docking analysis on wt-PfDHFR with and without the TS domain. Since chains A and B represented the DHFR domain while chains C and D represented the TS domains of both mutants, chain A was chosen accordingly. Wild-type PfDHODH (PDB ID: 1TV5)  was also selected. This complex was made up of the PfDHODH main protein chain, complexed with a sulfate ion, flavin mononucleotide (FMN), orotic acid, pentaethylene glycolmonooctyl ether, and an inhibitor A771726 (teriflunomide). Protein preparation was achieved by removal of all water molecules and complexed ligands of the proteins from their respective binding pockets, addition of polar hydrogens, and calculation of Gasteiger charges for each receptor atom to make them suitable for docking using AutoDockTools-1.5.7rc1 .
2.2. Ligand Preparation and Physicochemical Classification
The structures of the alkaloids and other ligands used in this study are presented in Figure 2. Three-dimensional (3-D) builds of title ligands were generated and optimized quantum mechanically, as presented in the work of Borquaye and coworkers . Polar hydrogens and Gasteiger charges were added to each and thereafter written into their pdbqt formats. Since ADME properties including physicochemical, pharmacokinetics, drug-likeness, and medicinal chemistry friendliness of the alkaloids have been reported earlier , structural classification based on quantum mechanical estimation of polar surface area (PSA, Å2), Corey–Pauling–Koltun (CPK) area (Å2), and volume (Å3) in Spartan’14 (Wavefunction Inc., Irvine California, USA) was performed. Distribution coefficient (logD, pH = 7) of the ligands was also estimated using Chemaxon (https://disco.chemaxon.com/calculators/demo/plugins/logd/).
2.3. Protein-Ligand Docking
Molecular docking experiments were performed with AutoDock Vina  extended into UCSF Chimera . Validation of docking protocols was made by redocking ligands (inhibitors) from the protein-ligand complexes into their binding sites. In order to verify docked poses as representative of correctly bound ligand conformations, overlays were created with experimentally determined binding modes and root mean square deviations (RMSD) estimated in PyMOL. The bound conformations of the redocked ligands and an initial blind docking influenced the centers and sizes of grid box coordinates (Table 1) of the respective binding pockets chosen. The poses of the ligands for each target were ranked according to their binding energies by AutoDock Vina in UCSF Chimera. Ligands with very low binding energies and optimal interactions with the active site residues were selected as the best poses [33, 34].
2.4. Molecular Dynamics Simulations
2.4.1. Computational Details
Molecular dynamics (MD) simulations were carried out to assess the probability of preservation of binding poses obtained from molecular docking studies and probe ensuing intermolecular interactions. To do this, selected protein-ligand complexes from molecular docking studies were used as starting structures. Complexes for MD simulations were sampled to obtain a good coverage of shape-size complementarity in binding of ligands of both ligand chemotypes (based on ligand classification), to corresponding protein targets. Simulations were performed using GROMACS v.2018.6  package on the Lengau cluster (Center for High Performance Computing, Cape Town). Ligand topologies were obtained by a CGenFF force field server, while protein topologies were generated using gmx pdb2gmx tool with the CHARMM36 all-atom force field. Generated complexes were solvated with a TIP3-P water model in a dodecahedral box type. System-specific neutralizations were carried out by addition of specific numbers of Na+ and Cl− counter ions. Energy minimizations were achieved by the CHARMM36 force field without constraints using the steepest descent algorithm for 50,000 steps, until a tolerance of 10 kJ/mol, for each system. After equilibration at 300 K using the V-rescale (modified Berendsen thermostat), for 100 ps under constant number, volume, and temperature (NVT) conditions, the system was then subjected to constant number, pressure, and temperature (NPT) equilibration at 1 atm pressure for 100 ps. The linear constraint solver (LINCS) algorithm was used for positional constrain. The Particle Mesh Ewald (PME) method was used to compute long-range electrostatic interactions. Coulomb and van der Waals cutoffs were set to 1.2 nm. The defined time step for the production run was 2 fs, writing coordinate trajectories every 10 ps. Three-dimensional (3D) periodic boundary conditions (pbc) were set in all cases followed by a production run of 50 ns [36, 37].
2.4.2. Input Checks and Trajectory Analysis
Prior to the production run, initial quality checks on system total pressure and temperature and potential and kinetic energies were monitored to ensure good simulation quality. Trajectories were corrected to conform to periodic boundary conditions using gmx trjconv tool. The root mean square deviation (RMSD), root mean square fluctuation (RMSF) of residues, and radius of gyration (Rg) were analyzed using standard GROMACS tools (rms, rmsf, and gyrate) .
2.4.3. Binding and Residue Contribution Energies
Thermodynamics of binding were assessed for all frames using the Molecular Mechanics Poisson–Boltzmann Surface Area (MM-PBSA) method implemented in GROMACS through the g_mmpbsa module . The average binding energies (kJ/mol) of protein-ligand complexes in the solvent were estimated from computed molecular mechanics (EMM) and polar and nonpolar solvation free energy (Gsolv) energy terms. For calculations of polar solvation energy (Gpolar), g_mmpbsa relies on the Assisted Poisson–Boltzmann Solver (APBS) program. For nonpolar solvation energy (Gapolar) calculations, the Solvent Accessible Surface Area (SASA) model was used. EMM or vacuum potential energy consisted of electrostatic contribution (Eelec) and van der Waals (EvdW) contributions. Available python scripts (MmPbSaDecomp.py and MmPSaStat.py) were used for these estimations, including residue-wise free energy contributions (theoretical details provided elsewhere [37, 38]), which were subsequently converted into corresponding β-factor using the energy2bfac module, for visualization.
2.4.4. Hydrogen Bonding Analysis, Visualization, and Plotting
The nonbonding interactions ensuing within protein-ligand complexes written from UCSF Chimera after molecular docking and specific frames extracted from MD ensembles were analyzed with the Discovery Studio 2017 R2 client (BIOVIA Discovery Studio Visualizer, Dassault Systèmes, San Diego). Visualization of MD trajectories and hydrogen bonding analysis was achieved in VMD 1.9.3 , using the produced structural (.gro) and pbc corrected compressed trajectory (.xtc) files of all complexes considered. A stride of 2 was set to give 2500 frames for hydrogen bonding analysis and MD-movie production. All hydrogen bonds were evaluated as well as polar hydrogen bonds with angle and distance cutoff set to 180° and 3.0 Å, respectively, to give the number of hydrogen bonding interactions as well as their occupancies. Resulting molecular graphics were generated from the same. Datasets generated were plotted in XMGRACE  and MS Excel.
3.1. Physicochemical Classification
C. sanguinolenta alkaloids were grouped by their CPK area and volume. Spatially bulky (sb) chemotype described compounds with two indoloquinoline moieties. Cryptolepicarboline, cryptoquindoline, cryptomisrine, biscryptolepine, and cryptospirolepine were classified as belonging to the sb chemotype as a result. The compounds in the sb chemotype group had CPK values ranging between 400.39 and 491.90 Å3. The planar (pg) chemotype class included cryptolepine, neocryptolepine, isocryptolepine, cryptoheptine, cryptolepinone, hydroxycryptolepine, and 11-isopropylcryptolepine. These compounds possessed a single indoloquinoline unit and had CPK volumes between 246.05 and 299.80 Å3. The structural classification of the alkaloids of C. sanguinolenta is shown in Figure S1. The calculated PSA for all alkaloids ranged between 7.01 and 37.31 Å2. The observed PSA trend grouped 11-isopropylcryptolepine, neocryptolepine, isocryptolepine, cryptolepicarboline, and cryptolepine together, with relatively lower areas (7.01 to 8.99 Å2). Hydroxycryptolepine, cryptolepinone, cryptospirolepine, cryptoheptine, and cryptomisrine possessed much higher PSA values from 27.10 to 37.31 Å2 with cryptoquindoline (15.44 Å2) and biscryptolepine (16.01 Å2) having similar PSA values (Figure S1).
The distribution coefficients (logD, pH = 7) estimated for pyrimethamine and teriflunomide were 1.84 and 0.61, respectively. The logD of the pg class of alkaloids ranged between 2.49 (for cryptolepine) and 5.10 (11-isopropylcryptolepine), indicating moderate hydrophobicity. For the representative alkaloids of the sb class (biscryptolepine and cryptomisrine), the logD was >7.00 (Table S1), which is indicative of high hydrophobicity.
3.2. Validation of Docking Protocol
Validation of docking protocols was performed by docking the co-crystallized ligands with their respective proteins (pyrimethamine for wt-PfDHFR, dm-PfDHFR, WR99210 for qm-PfDHFR, and teriflunomide for PfDHODH) to verify the accuracy of predictability of experimental conformation based on set criteria. Evaluation of docking protocols was by inspection of docked ligand conformations, superimposition of poses with their experimental conformers, and RMSD computations (Figure 3). The deviations (RMSD) of the re-docked pyrimethamine from the cocrystallized pyrimethamine were 0.148 Å and 0.046 Å for wt-PfDHFR and dm-PfDHFR, respectively, and 0.773 Å for the WR99210 and qm-PfDHFR complex, while re-docked teriflunomide deviated by 0.960 Å from the crystallized version in PfDHODH. RMSD values less than 2, in general, represent good reproduction of the co-crystallized structure (Trott and Olson, 2010). The active site residues of wt-PfDHFR from its PDBSUM entry are Ile14, Cys15, Ala16, Asp54, Met55, Phe58, Ile164, Tyr170, and Thr185, and these residues are required for pyrimethamine binding (Figure 4). The results from the re-docking of pyrimethamine with the wt-PfDHFR indicated the same residues mediating pyrimethamine binding, as shown in Table 2. Hydrogen bonding interactions were observed between pyrimethamine and Ile14, Cys15, Asp54, and Thr185. Hydrophobic interactions between pyrimethamine and Ala16, Met55, Phe58, and Ile164 were also observed in both the experimental and re-docked poses in wt-PfDHFR and dm-PfDHFR, where mutated residues maintained similar interactions with the inhibitor (Tables 2 and 3). Similarly, the interactions of WR99210 with qm-PfDHFR were reproduced to a large extent (Table 4). Additionally, re-docked teriflunomide pose and interactions were consistent with those obtained from experiment (Table 5). The estimated deviations and interactions validated the computational choices and docking algorithms.
3.3. Protein-Ligand Binding
The binding energies of the alkaloids towards wt-PfDHFR (3QGT) ranged from −8.3 to −13.4 kcal/mol, and these represent higher affinities when compared to pyrimethamine (−8.0 kcal/mol) (Table 2). The binding affinity of pyrimethamine for dm-PfDHFR (1J3J) was −7.9 kcal/mol, whereas the alkaloids were bound to the same protein with binding energies ranging from −7.9 kcal/mol to −12.5 kcal/mol (Table 3). For WR99210, the cocrystallized ligand for the quadruple mutant, a binding affinity of −8.4 kcal/mol was obtained from the docking studies. The binding energies of the alkaloids against qm-PfDHFR were between −6.0 and −12.9 kcal/mol (Table 4).
Amino acid residues of PfDHFR that interacted with the C. sanguinolenta alkaloids included residues 16, 40–46, 54–58, 108–113, and 164–170 (Tables 2–4). In addition to these residues, cryptospirolepine had extra interactions with Asp194 and Val195 in its binding to dm-PfDHFR (Table 3). Alkaloids of sb chemotypes interacted relatively well with all variants of PfDHFR, with binding energies less than −10.7 kcal/mol. Cryptospirolepine was however an exception, with binding energies of −8.8 kcal/mol, −9.9 kcal/mol, and −6.0 kcal/mol for wt-, dm-, and qm-PfDHFR, respectively. Alkaloids of the pg chemotype possessed less affinity for the DHFR proteins in comparison to their sb counterparts. The binding energies of the pg compounds ranged between −8.3 and −8.8 kcal/mol for wt-PfDHFR, −7.9 to −8.7 kcal/mol for dm-PfDHFR, and −8.6 to −9.1 kcal/mol for qm-PfDHFR.
The major bond types that mediated the interaction of pyrimethamine with wt- and dm-PfDHFR were hydrogen bonds to Ile14, Cys15, Asp54, and Thr185 and then hydrophobic interactions with Ala16, Met55, Phe58, and Ile164 (Figure 4). Cryptolepine and hydroxycryptolepine interacted with Asp54 of wt- and dm-PfDHFR via conventional hydrogen bonding. The interaction of Asp54 with cryptolepine involved the carboxylate oxygen of the amino acid and the amino hydrogen of cryptolepine. For hydroxycryptolepine, the interaction was between the same carboxylate oxygen of Asp54 and the hydroxyl hydrogen of the alkaloid (Figure 5).
Cryptoheptine was also observed to interact with Asp54 in both the wild type and double mutant variants of DHFR, whereas cryptolepinone interacted with Asp54 of the double mutant variant only (Figure 6).
Hydrophobic interactions mediated the binding of the other alkaloids with wt- and dm-PfDHFR. Strong hydrophobic interactions such as pi-alkyl interaction with Ala16, Val40, Val45, Leu46, and Ser111–Leu119 and in some cases pi-sulfur with Met55 and pi-pi stacking with Phe58 were mostly observed. No interactions between the alkaloids and Asp54 in the quadruple mutant were observed for all the alkaloids. Rather, interactions with the mutated Asn108 via hydrogen bonding were most prevalent.
The binding energy of teriflunomide in PfDHODH was determined to be −9.8 kcal/mol. The binding energies of alkaloids of C. sanguinolenta against PfDHODH ranged between −8.2 and −10.6 kcal/mol (Table 5). Teriflunomide was observed to bind in the quinone-binding tunnel, making important interactions with Cys184, His185, Phe188, Arg265, Tyr528, Val532, Gly535, and Met536. Examination of the individual binding poses from the docking of the pg class of alkaloids with PfDHODH revealed that about 50% of the population were bound in the inhibitor binding pocket (same pocket where teriflunomide binds), with the other 50% binding in the region occupied by the FMN cofactor, herein referred to as the FMN binding domain (FMN-BD).
The binding energies of the pg class of alkaloids in both the inhibitor and the FMN binding domains are provided in Table S2. For poses that were bound in the inhibitor binding domain (IBD), interactions with Cys184 and Val532 were observed for all alkaloids of the pg class (Table 5). Most of the pg alkaloids also interacted with Met536. Cryptoheptine and hydroxycryptolepine formed hydrogen bonds with the imidazolyl N1 of His185, using their oxo-groups at distances of 2.59 and 2.79 Å, respectively (Figure 7). In addition to this, cryptoheptine had polar contacts with Cys184 at a distance of 2.8 Å. The quinolyl cyclic keto group and pi-orbitals in cryptolepinone also formed strong polar and pi-donor hydrogen bonds at 3.17 Å and 3.37 Å, respectively, with Cys184. Cryptolepine and neocryptolepine had binding energies of −10.6 kcal/mol, followed by cryptolepinone and isocryptolepine (−10.0 and −10.2 kcal/mol, respectively), with strong hydrophobic interactions with the residues in the inhibitor binding site. Interestingly, alkaloids of the sb class did not interact with any of the residues in the inhibitor binding domain or the FMN binding domain. These alkaloids interacted with PfDHODH outside the quinone-binding tunnel (Figure 8(c)). Alkaloids of the sb class mostly interacted with residues such as Leu176, Leu177, Tyr178, and Arg265 along the edge of the quinone binding tunnel of PfDHODH via pi-cation and strong hydrogen bonding (1.98–3.06 Å) (Figure 8). This space was, however, solvent accessible and is depicted with blue shades around ligands and solvable residues in the figure.
3.4. Molecular Dynamics Simulations
Since docking neglects conformational changes that occur in the protein, inclusion of solvation, ionic effects, entropy, and ensuing force fields in an ensemble that contribute to binding affinities, the binding energy of a ligand could be overestimated. MD simulations combined with free energy analysis and other molecular descriptor evaluations were used to further interrogate the binding event over a 50 ns timescale. For wt- and dm-PfDHFR, alkaloids selected for MD simulation were biscryptolepine, cryptomisrine, and cryptospirolepine (for the sb class) and cryptolepine, cryptoheptine, cryptolepinone and hydroxycryptolepine (for the pg class). The same alkaloids, in addition to cryptolepicarboline (sb) and neocryptolepine (pg) were used for the qm-PfDHFR. Biscryptolepine (for the sb class), cryptolepine, cryptoheptine, cryptolepinone, isocryptolepine, neocryptolepine, and hydroxycryptolepine (all for the pg class) were used when PfDHODH was considered.
3.5. Stability of Protein-Ligand Complexes
The relative average deviations of protein backbone residues from the PDB crystal structures of the PfDHFR variants were 0.398 nm for wt-PfDHFR (Figure 9(a)), 0.354 nm for dm-PfDHFR (Figure 10(a)), and 0.398 nm for qm-PfDHFR (Figure 11(a)), and that of PfDHODH was 0.376 nm (Figure 12(a)). Postequilibration RMSDs (indicated as red lines in Figures 9(a), 10(a), 11(a), and 12(a)) were estimated relative to the first MD frame, for each of the systems considered. The average deviation of dm-PfDHFR from its starting frame was 0.270 nm, and that for qm-PfDHFR and wt-PfDHFR were 0.326 nm and 0.331 nm, respectively. The average deviation of PfDHODH from its starting frame was 0.200 nm. The RMSD of ligands with reference to their respective protein backbones were used to evaluate the preservation of binding poses obtained from the molecular docking in Vina. The stability of ligands under dynamic conditions was observed to be influenced by the overall strength of the protein-ligand interactions found in each protein-ligand system. For wt-PfDHFR (Figure 9), ligand conformations over the period of simulation were conserved for cryptolepine, cryptoheptine, and hydroxycryptolepine (all in the pg class). The other ligand in the pg class, cryptolepinone, exhibited significant conformational deviations after 3 ns of the MD simulation. This new conformation was averagely maintained till ∼26 ns. Another major conformational shift was observed at this time, and the structure persisted to the end of the run. Even though there were deviations, all the conformations of cryptolepinone were within the active site pocket of wt-PfDHFR. Cryptomisrine and cryptospirolepine (sb class) exhibited relatively stable conformations in complex with wt-PfDHFR, whereas for biscryptolepine (also of the sb class), deviations were observed after 5 ns. The biscryptolepine conformation after 5 ns was maintained till the 50th ns. Cryptolepine and hydroxycryptolepine (both pg class), as well as biscryptolepine (sb class), exhibited very little deviation from the starting structures when bound to dm-PfDHFR (Figure 10). For cryptolepinone in complex with dm-PfDHFR, significant deviations were observed after 36 ns. For cryptoheptine in complex with the same dm-PfDHFR protein, deviations were seen 3 ns into the simulation, and this stabilized for the duration of the entire experiment (Figure 10). All pg type ligands in complex with qm-PfDHFR were relatively stable for the whole 50 ns simulation period (Figure 11). For the sb class ligand–qm-PfDHFR complexes, biscryptolepine, cryptolepicarboline, and cryptospirolepine behaved similarly as their pg counterparts, with stable structures observed throughout the entire simulation period. For cryptomisrine, a significant deviation was observed after 13 ns, and this conformation was maintained till 45 ns. Thereafter, a further deviation was observed till 50 ns. For the PfDHODH-ligand complexes (Figure 12), all the pg class of ligands maintained stable conformations during the MD runs. The biscryptolepine-PfDHODH complex deviated significantly after 4 ns and then achieved stability thereafter till the end of the simulation.
To interrogate residue fluctuations, the RMSF of alpha carbons were estimated. In general, high fluctuations were observed in the loop regions of the various proteins studied, as expected. Unusually high fluctuations were observed for biscryptolepine, cryptomisrine, cryptospirolepine (all sb ligands), and cryptoheptine (pg ligand) complexes with the PfDHFR variants. Regions where these profound fluctuations were observed include residues 20–40, 75–100, and 130–140 (Figures S2(a), S3(b), and S4(b)). In both wt- and dm-PfDHFR, these fluctuations were profound. Surprisingly, similar fluctuations were absent in qm-PfDHFR. In the PfDHODH complexes, fluctuations observed were minimal, probably due to the absence of lengthy loops in the crystal structure (Figures S5 and S6). Interactions induced by cryptoheptine binding resulted in fluctuations around residues 340–350 and 475–80 (Figure S5(a)). Overall crystal packing was similar for all dynamic systems of PfDHFR variants studied. A general convergence of radii of gyration (Rg) between 1.80 nm and 1.95 nm were observed for all PfDHFR systems studied (Figures S2–S4). For the PfDHODH systems, Rg was between 1.90 and 2.10 nm (Figures S5 and S6).
3.6. Endpoint Binding (MMPBSA) Profiles
The MMPBSA method was used to estimate binding free energies (ΔGbind) of the protein-ligand complexes studied. For all alkaloids selected for MD analysis, nonpolar contributions (van der Waals and SASA) to binding were more pronounced. This is evidenced by the large negative ΔEvdW values recorded for the complexes. The wt-PfDHFR complexes with all sb ligands recorded spontaneous binding, with ΔGbind for cryptomisrine, biscryptolepine, and cryptospirolepine ranging between −85.79 and −130.82 kJ/mol. For the pg ligands, ΔGbind for cryptoheptine and cryptolepinone were negative, but that of hydroxycryptolepine (163.87 kJ/mol) and cryptolepine (188.37 kJ/mol) were positive which implies unfavorable overall binding. The binding free energies for both biscryptolepine and cryptomisrine were negative, whereas that of cryptospirolepine was positive (20.93 kJ/mol) for their complexes with dm-PfDHFR. The dm-PfDHFR-hydroxycryptolepine and dm-PfDHFR-cryptolepine complexes exhibited positive free binding energies, whereas those of the dm-PfDHFR-cryptoheptine and dm-PfDHFR-cryptolepinone complexes were negative. In the qm-PfDHFR complexes with biscryptolepine, cryptomisrine, and cryptospirolepine (sb class), as well as cryptoheptine and cryptolepinone (pg class), all recorded spontaneous negative binding free energies. The rest of the ligands investigated with qm-PfDHFR, however, gave positive free binding energy. High solvation potential (low ΔGsolv value) with electrostatic energy compensation, generally led to overall favorable binding in the PfDHFR variants. It is interesting to note that binding was spontaneous (negative ΔGbind; kJ/mol) for cryptomisrine, biscryptolepine, cryptoheptine, and cryptolepinone over the entire simulation period, for all protein-ligand complexes (Tables 6 and 7). The binding affinities of biscryptolepine and cryptomisrine towards the PfDHFR proteins were in the order dm-PfDHFR > qm-PfDHFR > wt-PfDHFR. For cryptoheptine, the binding affinities decreased with increasing DHFR mutations (i.e., wt-PfDHFR > dm-PfDHFR > qm-PfDHFR), whereas the order was dm-PfDHFR > wt-PfDHFR > qm-PfDHFR for cryptolepinone (Table 6).
As indicated earlier during the analysis of the molecular docking data, the pg class of alkaloids occupied both the inhibitor and FMN binding domains of PfDHODH equally. Therefore, the systems used in the MD simulations reflected these observations. For binding in both the inhibitor and FMN binding domains of PfDHODH over the entire 50 ns simulation period, cryptoheptine and cryptolepinone recorded overall negative free binding energies (−115.28 and −112.09 kJ/mol, respectively, for IBD bound complexes; −89.85 and −86.03 kJ/mol, respectively, for FMN-BD bound complexes). All the other pg class of alkaloids examined (cryptolepine, isocryptolepine, neocryptolepine, and hydroxycryptolepine) over the same 50 ns period exhibited positive binding energies, ranging from 9.31 kJ/mol to 27.43 kJ/mol for IBD-bound complexes and 32.22 kJ/mol to 64.06 kJ/mol for FMN-BD bound complexes. Careful examination of the last 15 ns of the MD simulation of the pg class of alkaloids bound in both binding domains of PfDHODH revealed improved binding affinities for all the alkaloids. Cryptolepine, isocryptolepine, and hydroxycryptolepine, which had positive overall binding energies for the entire 50 ns simulation, exhibited negative binding energies in the last 15 ns for binding in both the IBD and FMN-BD. Neocryptolepine was however an exception. When bound in the FMN-BD, positive binding energies were recorded for both the entire 50 ns simulation and the last 15 ns extraction. For binding in the IBD, the overall binding energy (50 ns simulation) was positive, but this energy was negative in the last 15 ns of the simulation. Biscryptolepine (sb class) did not dock at either the IBD or FMN-BD, but rather outside the quinone-binding tunnel. Thus, the estimated thermodynamic parameters could be regarded as a negative control case.
3.7. Hydrogen Bonding and Other Nonbonding Interactions
Estimation of total number of hydrogen bonds made between ligands and proteins over the entire simulation revealed that all systems that exhibited negative ΔGbind values engaged in at least 2 hydrogen bonds. The highest hydrogen bond count of all the systems studied were 10 and 7 for cryptospirolepine and cryptomisrine, respectively, when both ligands were complexed with wt-PfDHFR, and this occurred at about halfway into the simulation. The highest hydrogen bond contacts by the ligands were with water molecules, rather than with any amino acid residues (Figures S7–S9 (i), (ii), and (iii)). Biscryptolepinone, cryptomisrine, cryptospirolepine, and cryptoheptine all made hydrogen bond contacts with 3 residues in wt-PfDHFR during the MD simulation. All 4 ligands interacted with Ser111. Biscryptolepine, cryptomisrine, and cryptoheptine all interacted with Ser108 as well. The occupancies (considering both side chains and main protein backbone) for cryptomisrine-Ser108 and cryptospirolepine-Ser111 were all greater than 90 (Table S3). Hydroxycryptolepine (a pg ligand with a positive MMPBSA binding free energy) recorded a hydrogen bond occupancy of 85% with Asp54, a key residue involved in the inhibition of wt-PfDHFR by pyrimethamine (Table S3). In dm-PfDHFR, biscryptolepine, cryptomisrine, and cryptospirolepine all interacted with both Asn108 and Ser111. The cryptomisrine-Ser111 H-bond occupancy was very frequent, at 61.89%. Only cryptolepinone occupied Asp54 (30.97%). For cryptoheptine, its interaction with Phe116 was very strong, with an occupancy >80%. Once again, hydroxycryptolepine displayed strong affinity towards Asp54, with an occupancy of both the side chain and protein backbone recorded at about 200% (Table S4). Hydroxycryptolepine also interacted with Asn108, even though it was only to a minimal extent (1.2%). H-bond occupancy with respect to qm-PfDHFR was relatively lower than that for both wt-PfDHFR and dm-PfDHFR. The highest occupancy (9.08) was recorded by the interaction of Asn108 with biscryptolepine. Once again, biscryptolepinone, cryptomisrine, and cryptospirolepine all interacted with Ser111 (Table S5).
Deviations observed for cryptolepinone-qm-PfDHFR complex after 25 ns (Figure S4(a)) were observed to have been mediated by weak interactions with Cys50 and Met55, however, strong with water molecules using its carbonyl oxygen (Figures S9 (ii), B1 and B2). Cryptolepinone unbinding after ∼28 ns from qm-PfDHFR binding pocket was mediated by strong hydrogen bonding with solvent residues (Figures S9 (ii), B1 and B2). This led to a complete displacement from the protein binding site. Cryptospirolepine conformations predicted from docking were stable for wt- and qm-PfDHFR but not for the double mutant. Visual inspection of cryptospirolepine starting conformation revealed weak interactions with Asp194, Val195, Gly44, and unfavorable interactions with its carbonyl oxygen, coupled with high solvent accessibility (Figure S3 A1), could account for cryptospirolepine unbinding over the 50 ns simulation. Very strong hydrogen bonding interactions (conventional and carbon-hydrogen types) existed for Leu40, Gly44, Leu46, Asp54, Arg59, and Ser108 or Asn108, Phe116, Leu164, and Tyr170 for both classes of alkaloids interacting with the DHFR variants. These re-enforced respective nonpolar interactions. Based on the strength of nonpolar interactions, these hydrogen bonding interactions were maintained up to the end of the simulation (Figures S7–S9 (i), (ii), and (iii)). This was obvious for cryptomisrine-wt-PfDHFR binding at both 30 ns and 50 ns (Figures S7 (i)) where SOL8700 contributed significantly to ligand binding.
For PfDHODH, cryptoheptine and cryptolepinone of the pg class of alkaloids interacted well with Gly181, His185, Val532, and Phe227 in the inhibitor binding domain, with polar hydrogen bond occupancies ranging from 1.32 to 21.08%. In their interaction with the FMN-BD, cryptoheptine formed H-bonds with Gly69, Lys72, Asn117, Ser282, Ser310, and Try333 with a bond occupancy range of 2.56–41.38%. FMN-BD residues that interacted with cryptolepinone are Gly66 (12.32%), Ala68 (8.88%), Thr92 (18.00%), and Asn122 (14.71%) (Table S6). Regarding this observation, though residue contributions for cryptolepinone binding were generally strong, cryptoheptine produced high selectivity for both the IBD and FMN-BD. Phe17, Phe14, and Leu10 contributed the most to biscryptolepine binding to the IBD with residue contributions of −9.63 kJ/mol (Figures 12 and S7). Also, biscryptolepine interacted weakly with residues around the end of the quinone-binding tunnel, hence its displacement by water to the mouth region and finally to the interface of the waist region (Figure 12(a)), where binding was favorable for the remaining 38 ns. This was because alkaloids of the sb class were bound around solvent accessible sites in similitude of the case of cryptospirolepine (Figure S3 A1), though bound around this region, high affinity was obtained from nonpolar interactions made with Val45, Cys50, Met55, and Pro113 in the quadruple mutant.
The indolo[3,2-b]quinoline alkaloids of C. sanguinolenta have been shown to have a broad spectrum of pharmacological effects in multiple disease states . Unfortunately, like most bioactive compounds, a lack of adequate characterization of specific biomolecular targets with which they interact to elicit observed biological activity impedes drug development. Accordingly, this work sought to elucidate the potential interactions of C. sanguinolenta alkaloids with druggable proteins (DHFR and DHODH) involved in the folate biosynthetic network of Plasmodium falciparum . For PfDHFR, variants with 2 and 4 important amino acid mutations have contributed significantly to pyrimethamine resistance. Thus, in addition to the wild-type variant, both double and quadruple mutant variants were also studied. The molecular docking study was carried out to study the interactions between the ligands and the selected proteins and also deduce the most energetically favorable conformation of the ligand upon binding as a starting complex for molecular dynamics (MD) simulations. MD simulations provided atomistic details ensuing under dynamic conditions for 50 ns. Also, estimation of free energies was carried out for scoring and ranking purposes. AutoDock Vina is better at predicting binding conformations with higher accuracy than estimating binding energies . MMPBSA was therefore used to estimate the end-point binding free-energies from the MD ensembles obtained after the simulations . Analysis of nonbonding interactions was used to identify amino acids that are important for ligand recognition and evaluate the affinity of the ligands towards each protein target. In all optimal interactions with important residues within the binding region of each target, the stability of the complexes formed and estimated ΔGbind were used as criteria to predict inhibitory activity .
Yuvaniyama and coworkers have shown that Ile14, Cys15, Ala16, Asp54, Phe58, Asn108, Pro113, and Ile164 of wt- and dm-PfDHFR make important interactions with pyrimethamine and are thus necessary for inhibitory action. They have also shown that Ile14, Cys15, Asp54, Met55, Phe58, Leu119, Ile164, and Tyr170 of qm-PfDHFR do interact with WR99210, which is a known inhibitor . The alkaloids studied in this work were observed to make important interactions with Ala16, Asp54, Met55, Phe58, Ser108, Asn108, Ile112, Pro113, and Ile164 of the wild-type and double mutant variants of PfDHFR. Visual inspection of the binding modes also demonstrated that most of the ligands formed hydrogen bond interactions with at least one of the key amino acid residues (Asp54, Ile14, Leu/Ile164, and Asn/Ser108) and hydrophobic interactions with the hydrophobic residues (Phe58, Met55, Ile112, and Pro113) within the active sites of either the wild-type, double, or quadruple mutant variants of PfDHFR. High stabilities observed in MD trajectories of alkaloids of both ligand groups in most cases confirmed the accuracy of pose predictions, validating molecular docking outcomes.
Asp54 has been identified as a critical active site residue of PfDHFR involved in hydrogen bonding with inhibitors through its carboxylate oxygen atoms [4, 27, 46]. For all variants of PfDHFR, hydrogen bonding interactions between each of the cocrystallized ligands (WR99210 for qm-PfDHFR and pyrimethamine for both wt- and dm-PfDHFR) and Asp54 were observed (Figure 4). Recently, the reaction surface of DHFR has been explored with QM/MM tools and it has been shown that Asp54 is essentially responsible for hydride transfer from the NADPH cofactor to dihydrofolate (DHF), leading to the formation of tetrahydrofolate (THF), through π-stacking anchoring provided by Phe58 . Thus, by interacting with Asp54 and Phe58, inhibitors such as pyrimethamine, cycloguanil, and WR99210 interfere with hydride transfer, hence terminating downstream events that lead to effective DHF reduction.
Plasmodium parasites have a track record of developing resistance to virtually all available drugs. With respect to DHFR, resistance occurs as a result of mutations to one or more amino acid residues in the active site of the enzyme in the parasite. Cys59Arg + Ser108Asn and Asn51Ile + Cys59Arg + Ser108Asn + Ile164Leu are commonly observed mutations which are responsible for resistance. Amino acid residue, Ser108, in PfDHFR enzyme is located at the opposite end of the binding site to Asp54 and directly faces the active site cavity. Mutation of this residue is implicated in resistance development. Resistance is further enhanced by mutations at Asn51Ile and Cys59Arg. Asn51Ile mutation alters the location of the Asp54 carboxylate oxygens and the orientation of their H-bonds with the drug . These changes increase the volume of the active site or create steric clashes which result in displacement of inhibitors from their optimal orientation and prevents or weakens their interaction with Asp54, which is crucial for inhibitor binding within the active site and thereby aborting the antiplasmodial activity of current antifolates . The mutations present in dm-PfDHFR confer the most severe cases of antifolate resistance .
From the docking results, the alkaloids of the pg class interacted with either Asp54 or Phe58 of both wt- and dm-PfDHFR, suggesting that these alkaloids could likely interfere with DHF reduction in these variants and hence probably inhibit or significantly reduce enzyme activity (Tables 2 and 3). A closer examination of the trajectories obtained from MD simulation revealed that the bound conformations of the alkaloids of the pg class against wt-PfDHFR were largely unchanged. Even cryptolepine, which exhibited some significant orientational changes at ∼3 and ∼26 ns of the simulation was still bound in the active site pocket of the protein, as evidenced by interactions with Asp54 at 30 ns and 50 ns (Figure S7 (iii)). Binding of the ligands to both wt- and dm-PfDHFR did not stabilize the loop regions of the protein, and significant fluctuations were thus observed. However, the regions where these fluctuations occurred were outside the active site pocket and explains why the bound conformations observed in the docking study were maintained throughout the 50 ns MD simulation. The contribution of Phe58 to cryptoheptine binding, as determined from the MMPBSA binding free energy, was between −2 and −5 kJ/mol. The orientation of the aromatic rings of cryptoheptine was on top of the phenyl ring of Phe58 but oriented about 90° away (Figure 9(c)). For cryptolepinone, Phe58’s contribution to binding was between −0.5 and −1.33 kJ/mol. In the case of hydroxycryptolepine, the contribution of Phe58 was immense, with an energy contribution of −10.25 and −17.99 kJ/mol being observed. The average position of the aromatic ring of hydroxycryptolepine was directly on top of the phenyl ring, with interactions being largely hydrophobic (Figure 9(e)). In all cases, the interaction with Asp54 was insignificant. Thus, compounds of the pg class probably exert their inhibitory action by interacting with Phe58. In so doing, they reduce the capacity of Phe58 to anchor DHF for reduction to THF.
In the case of dm-PfDHFR, Asp54 did not make good residue energy contribution to both cryptoheptine and cryptolepinone binding (Figures 10(a)–10(d)). However, their interaction with Phe58 was important, with energy contributions ranging from −1.7 to −3.01 and −4.38 to −11.04 kJ/mol, respectively. The mode of inhibition is thus likely to be similar to that observed for all the alkaloids of the pg class in their binding to wt-PfDHFR. In the case of hydroxycryptolepine, however, the contributions of both Asp54 and Phe58 were immense. The energy contribution of Phe58 to hydroxycryptolepine binding was between −34 and −65 kJ/mol, whereas that of Asp54 was between −65.71 and −96.60 kJ/mol (Figure 10(e)). The aromatic ring of hydroxycryptolepine was directly stacked on top of the phenyl ring of Phe58. This suggests that both Asp54 and Phe58 have been engaged in binding with hydroxycryptolepine (Figure S7 (iii)), and hence these residues will be unable to participate effectively in hydride transfer and DHF anchoring—the two key processes involved in DHF reduction. This corroborates the data from the molecular docking experiment. For the quadruple mutant, both cryptoheptine and cryptolepinone did not interact significantly with Asp54, as was observed in the docking run (Figures 11(a) and 11(d)). Contribution of Phe58 to binding of both ligands was also weak. In the case of neocryptolepine, Asp54 contributed to its binding with an energy contribution of −20.18 kJ/mol, whereas Phe58 contributed an energy of −11.53 kJ/mol (Figure 11(e)). Neocryptolepine’s mode of inhibition is therefore likely to involve interruption in hydride transfer and anchoring of DHF during reduction.
Cryptomisrine (sb class) exhibited very good binding affinity towards wt-, dm-, and qm-PfDHFR in molecular docking (Tables 2–4). However, there was no interaction with either Asp54 or Phe58 in any of those complexes. A careful review of the contributions of these 2 residues to the overall ligand-protein stability in the MD simulation shows minimal influence (Figures 9(b), 10(b), and 11(b)). Thus, cryptomisrine probably could effect its inhibitory action on the DHFR variants by engaging the other residues (apart from Asp54 and Phe58), hence making DHF binding and subsequent reduction either impossible or much slower. Hydrophobic interactions with Ala16, Leu46, Met55, Phe58, Ile112, and Leu119 within the binding region of the PfDHFR variants contribute to overall stability of substrate binding , and hence interrupting some of these interactions could thwart protein activity. Data from RMSF computations showed that ligand binding to qm-PfDHFR largely reduced fluctuations in the loop regions of the protein. Despite its larger active site, there was very little fluctuations in the active site region of the protein when the ligands were bound. Together, these results suggest the formation of a stable protein-ligand complex throughout the simulation period.
Of the chemotypes in the pg class, though binding was predicted to be spontaneous in molecular docking, endpoint energy calculations yielded positive ΔGbind for all except cryptoheptine and cryptolepinone. However, concerning conservation of predicted poses, very strong hydrogen bonding interactions of cryptolepine and hydroxycryptolepine with Asp54 in both wt-PfDHFR and dm-PfDHFR were observed, giving an overall low conformational entropy. We therefore speculate that the MMPBSA estimation could not sample varying conformations to account for entropic contributions, as has been suggested elsewhere [44, 50, 51]. Since this intrinsic limitation exists for current tools used in estimating MMPBSA free-energies, other high level and computationally demanding approaches, such as free energy perturbation (FEP), may be used for future evaluations. Therefore, with binding energies estimated from docking and Asp54 residue contributions to ligand binding free energy, visual inspection of interactions over 30 ns and 50 ns suggests hydroxycryptolepine and cryptolepine as potential DHFR inhibitors by plausibly binding covalently to Asp54.
It has been suggested that potential PfDHFR enzyme inhibitors must fulfill at least three conditions to permit chemical and geometrical complementarity of the ligands with residues in the active site of PfDHFR. These are (i) presence of a hydrogen bond donor head group that can form favorable hydrogen bond interactions with Leu/Ile164, Asp54, and Ile14, (ii) a hydrophobic aromatic tail that will occupy the hydrophobic pocket of the active site where residues such as Phe58, Met55, Phe116, Pro113, and Ile112 reside, and finally (iii) a linker unit between the H-bond donor head groups and hydrophobic aromatic tail to guarantee flexibility, so as to avoid unfavorable steric clashes with Asn108 in the active site of the mutant PfDHFR [22, 52–54]. Although the alkaloids generally lack the aforementioned geometries in comparison to pyrimethamine analogues, they exhibited good binding affinities to the wild-type and double mutant variants of PfDHFR when compared to the bound ligands; hence, their mechanism of inhibition most likely differs from that proposed for pyrimethamine. Also, acquisition of resistance against these alkaloids will differ from those used by DHFR against pyrimethamine. This nonpyrimethamine-like mechanism of PfDHFR inhibition has also been observed for triclosan ; as well, the in vitro binding of some nonpyrimidine fragments hits whose chemotypes successfully cocrystallized with PfDHFR revealed a binding in the active site and the vicinity of catalytic residues .
The alkaloids were also docked against another important protein in the parasite PfDHODH (1TV5) where dual binding modes to both the FMN-BD and IBD were prevalent for the pg class , with high preference for the IBD. The inhibitor-binding site of PfDHODH, which is located close to the cofactor-binding site, is made up of two regions exhibiting high amphipathicity—a hydrogen bonding pocket which comprises His185, Tyr528, and Arg265 and a hydrophobic pocket where residues such as Met536 and Tyr168 are located . When superimposed into teriflunomide, it was observed that cryptoheptine, cryptolepinone, and hydroxycryptolepine were bound within the same binding region as the native ligands, as shown in Figure 7(b). Visual inspection of the poses of cryptolepinone, cryptoheptine, hydroxycryptolepine, and cryptolepine revealed interaction of the alkaloids with the hydrophobic residue Met536. Since hydrophobic interactions drive ligand binding to this tunnel and hydrogen bonding provides quinone specificity, all alkaloids of the pg class may be strong PfDHODH inhibitors exploiting full hydrophobic advantages in contrast to those of the sb class. Attempts at docking cryptospirolepine, biscryptolepine, cryptolepicarboline, and cryptoquindoline into the inhibitor binding region of the PfDHODH enzyme failed probably because of their overall larger size. The active site pocket in PfDHODH is about 55 Å3, whereas the size of the tunnel in human and rat DHODH is much larger at 715 Å3 and 760 Å3, respectively. The smaller overall size of the PfDHODH active site explains why inhibitors with larger sizes, as alkaloids of the sb class, could be ineffective inhibitors of PfDHODH . These sb alkaloids could be evaluated as potential drugs in ameliorating disease phenotypes mediated by human DHODH, such as psoriasis, autoimmune diseases, cancer, transplant rejection, and rheumatoid arthritis [58, 59].
Molecular docking predicted stable ligand binding conformations in general. Fluctuations observed in ligand trajectories are attributable to high solvation potential and large desolvation penalties. Hydrophobic interactions were crucial for molecular recognition following relatively high hydrophobicity of alkaloids investigated. Together, the results of the docking and molecular dynamics assessment of the Cryptolepis sanguinolenta alkaloids show that biscryptolepine and cryptomisrine (sb alkaloids), as well as cryptoheptine, cryptolepinone, cryptolepine, and neocryptolepine (pg alkaloids), are potential inhibitors of PfDHFR. Alkaloids of the pg class mostly preferred to bind to the IBD over the FMN-BD of PfDHODH primarily due to size considerations. Because natural extracts usually contain a mixture of compounds that may act in synergy, this work justifies a plausible multifaceted inhibition of key proteins in the Plasmodium falciparum parasite that may contribute to the potent polypharmacological efficacy observed for C. sanguinolenta alkaloids. Overall, this work provides a mechanistic basis for the antiplasmodial action of C. sanguinolenta extracts and its isolated alkaloids.
All data generated or analyzed during this study are included in this published article.
Conflicts of Interest
All authors declare that there are no financial, professional, or personal conflicts of interest that might have influenced the performance or presentation of this study.
LSB conceived the study. All experiments were designed by LSB, LKK, GBA, JOM, and ENG. Computations were made by LKK, ENG, JOM, and GBA. Data analysis was done by LKK, ENG, JOM, GBA, and LSB. The manuscript was prepared by LKK, ENG, JOM, GBA, and LSB. All authors read and approved the final manuscript.
The authors are grateful to the Center for High Performance Computing, Cape Town, South Africa, who granted the authors generous access to the Lengau cluster for the MD simulations.
Table S1: distribution coefficient of C. sanguinolenta alkaloids expressed as LogD. Table S2: binding affinities determined for the formation of PfDHODH-ligand complexes from precision (IBD) and blind docking (SBD) experiments. Table S3: summary of hydrogen bond occupancies and residues involved for the wild-type P. falciparum dihydrofolate reductase (wt-PfDHFR) observed for ligands with favorable (negative) MMPBSA binding free energies based on set criteria (angle = 180° and distance = 3.0 Å). Table S4: summary of hydrogen bond occupancies and residues involved for the double mutant P. falciparum dihydrofolate reductase (dm-PfDHFR) observed for ligands with favorable (negative) MMPBSA binding free energies based on set criteria (angle = 180° and distance = 3.0 Å). Table S5: summary of hydrogen bond occupancies and residues involved for the quadruple-mutant P. falciparum dihydrofolate reductase (qm-PfDHFR) observed for ligands with favorable (negative) MMPBSA binding free energies based on set criteria (angle = 180° and distance = 3.0 Å). Table S6: summary of hydrogen bond occupancies and residues involved for the wild-type P. falciparum dihydroorotate dehydrogenase (PfDHODH) observed for ligands with favorable (negative) MMPBSA binding free energies based on set criteria (angle = 180° and distance 3.0 Å). Figure S1: classification of C. sanguinolenta alkaloids into groups based on structural features—CPK area and volume. Figure S2: overview of wt-PfDHFR-complex stability for different ligands and hydrogen bonding. Figure S3: trajectory analysis for cryptospirolepine bound to dm-PfDHFR and overview of complex stability and hydrogen bonding. Figure S4: trajectory analysis for cryptolepinone bound to qm-PfDHFR and overview of complex stability and hydrogen bonding interactions. Figure S5: overview of complex stability in different complexes in which ligands were bound in and around the quinone-binding tunnel of PfDHODH, as well as hydrogen bond count. Figure S6: summary of dynamic events contributing to complex stability in different complexes in which ligands were bound in and around the substrate FMN-binding pocket of PfDHODH, as well as hydrogen bond count. Figure S7: hotspot of interacting protein (PfDHODH) residues (colored by beta-factor) with starting MD conformations of binding partners: (A) biscryptolepine, (B) cryptoheptine, and (C) cryptolepinone. Figure S7 (i): visualization of protein-ligand interactions for 30 ns and 50 ns frames for biscryptolepine and cryptomisrine (sb class) bound to the wt-PfDHFR. Figure S7 (ii): visualization of protein-ligand interactions for 30 ns (left) and 50 ns (right) frames for cryptoheptine (A) and cryptolepinone (B) (pg class, with negative ΔGbind) bound to the wt-PfDHFR. Figure S7 (iii): visualization of protein-ligand interactions for 30 ns (left) and 50 ns (right) frames for cryptolepine (A) and hydroxycryptolepine (B) (pg class, with positive ΔGbind) bound to the wt-PfDHFR. Figure S8 (i): visualization of protein-ligand interactions for 30 ns and 50 ns frames for biscryptolepine (A) and cryptomisrine (B) (sb class) bound to the dm-PfDHFR. Figure S8 (ii): visualization of protein-ligand interactions for 30 ns and 50 ns frames for cryptoheptine (A) and cryptolepinone (B) (pg class, with negative ΔGbind) bound to the dm-PfDHFR. Figure S8 (iii): visualization of protein-ligand interactions for 30 ns (left) and 50 ns (right) frames for cryptolepine (A) and hydroxycryptolepine (B) (pg class, with positive ΔGbind) bound to the dm-PfDHFR. Figure S9 (i): visualization of protein-ligand interactions for 30 ns and 50 ns frames for biscryptolepine (A) and cryptomisrine (B) (sb class) bound to the qm-PfDHFR. Figure S9 (ii): visualization of protein-ligand interactions for 30 ns and 50 ns frames for cryptoheptine (A) and cryptolepinone (B) (pg class, with negative ΔGbind) bound to the qm-PfDHFR. (B) Figure S9 (iii): visualization of protein-ligand interactions for 30 ns (left) and 50 ns (right) frames for hydroxycryptolepine (A) and neocryptolepine showing only 30 ns frame (B) (pg class, with positive ΔGbind) bound to the qm-PfDHFR. (Supplementary Materials)
WHO, World Malaria Report 2018, World Health Organization, Geneva, Switzerland, 2019.
M. J. Gardner, N. Hall, E. Fung et al., “Genome sequence of the human malaria parasite Plasmodium falciparum,” Nature, vol. 419, no. 6906, pp. 498–511, 2002.View at: Publisher Site | Google Scholar
M. A. Phillips and P. K. Rathod, “Plasmodium dihydroorotate dehydrogenase: a promising target for novel anti-malarial chemotherapy,” Infectious Disorders-Drug Targets, vol. 10, no. 3, pp. 226–239, 2010.View at: Publisher Site | Google Scholar
Y. Yuthavong, J. Yuvaniyama, P. Chitnumsub et al., “Malarial (Plasmodium falciparum) dihydrofolate reductase-thymidylate synthase: structural basis for antifolate resistance and development of effective inhibitors,” Parasitology, vol. 130, no. 3, pp. 249–259, 2005.View at: Publisher Site | Google Scholar
J. E. Hyde, “Exploring the folate pathway in Plasmodium falciparum,” Acta Tropica, vol. 94, no. 3, pp. 191–206, 2005.View at: Publisher Site | Google Scholar
A. C. Anderson, “Targeting DHFR in parasitic protozoa,” Drug Discovery Today, vol. 10, no. 2, pp. 121–128, 2005.View at: Publisher Site | Google Scholar
V. Vyas and M. Ghate, “Recent developments in the medicinal chemistry and therapeutic potential of dihydroorotate dehydrogenase (DHODH) inhibitors,” Mini Reviews in Medicinal Chemistry, vol. 11, no. 12, pp. 1039–1055, 2011.View at: Publisher Site | Google Scholar
T. Rodrigues, R. Moreira, and F. Lopes, “New hope in the fight against malaria?” Future Medicinal Chemistry, vol. 3, no. 1, pp. 1–3, 2011.View at: Publisher Site | Google Scholar
J. Baldwin, A. M. Farajallah, N. A. Malmquist, P. K. Rathod, and M. A. Phillips, “Malarial dihydroorotate dehydrogenase,” Journal of Biological Chemistry, vol. 277, no. 44, pp. 41827–41834, 2002.View at: Publisher Site | Google Scholar
B. Hanboonkunupakarn and N. J. White, “The threat of antimalarial drug resistance,” Tropical Diseases, Travel Medicine and Vaccines, vol. 2, no. 1, p. 10, 2016.View at: Publisher Site | Google Scholar
F. Ayisi, C. N. Mensah, and L. S. Borquaye, “In vivo antiplasmodial activity and toxicological analyses of the ethanolic leaf and twig extract of faurea speciosa welw. (Proteaceae),” Journal of Parasitology Research, vol. 2021, Article ID 7347532, 12 pages, 2021.View at: Publisher Site | Google Scholar
M. K. Laryea and L. S. Borquaye, “Antimalarial efficacy and toxicological assessment of extracts of some Ghanaian medicinal plants,” Journal of Parasitology Research, vol. 2019, Article ID e1630405, 9 pages, 2019.View at: Publisher Site | Google Scholar
F. Ayisi, C. N. Mensah, and L. S. Borquaye, “Antiplasmodial potential and safety evaluation of the ethanolic stem bark extract of Distemonanthus benthamianus baill.(Leguminosae),” Scientific African, vol. 12, Article ID e00809, 2021.View at: Publisher Site | Google Scholar
G. Afu, M. K. Laryea, and L. S. Borquaye, “Biological efficacy and toxicological evaluation of ethanolic extract of Cassia nodosa buch.-ham.(leguminosae),” Journal of Chemistry, vol. 2020, Article ID 3983491, 13 pages, 2020.View at: Publisher Site | Google Scholar
M. K. Laryea and L. Sheringham Borquaye, “Antimalarial, antioxidant, and toxicological evaluation of extracts of Celtis africana, Grosseria vignei, Physalis micrantha, and Stachytarpheta angustifolia,” Biochemistry Research International, vol. 2021, Article ID 9971857, 10 pages, 2021.View at: Publisher Site | Google Scholar
WHO, National Policy on Traditional Medicine and Regulation of Herbal Medicines: Report of a WHO Global Survey, World Health Organization, Geneva, Switzerland, 2005, https://apps.who.int/iris/handle/10665/43229.
N. Osafo, K. B. Mensah, and O. K. Yeboah, “Phytochemical and pharmacological review of Cryptolepis sanguinolenta (lindl.) schlechter,” Advances in Pharmacological Sciences, vol. 2017, Article ID 3026370, 13 pages, 2017.View at: Publisher Site | Google Scholar
K. Cimanga, T. De Bruyne, L. Pieters, A. J. Vlietinck, and C. A. Turger, “In vitro and in vivo antiplasmodial activity of cryptolepine and related alkaloids from Cryptolepis sanguinolenta,” Journal of Natural Products, vol. 60, no. 7, pp. 688–691, 1997.View at: Publisher Site | Google Scholar
P. Grellier, L. Ramiaramanana, V. Millerioux et al., “Antimalarial activity of cryptolepine and isocryptolepine, alkaloids isolated from Cryptolepis sanguinolenta,” Phytotherapy Research, vol. 10, no. 4, pp. 317–321, 1996.View at: Publisher Site | Google Scholar
A. Paulo, E. T. Gomes, J. Steele, D. C. Warhurst, and P. J. Houghton, “Antiplasmodial activity of Cryptolepis sanguinolenta alkaloids from leaves and roots,” Planta Medica, vol. 66, no. 1, pp. 30–34, 2000.View at: Publisher Site | Google Scholar
A. D. Forkuo, C. Ansah, K. B. Mensah et al., “In vitro anti-malarial interaction and gametocytocidal activity of cryptolepine,” Malaria Journal, vol. 16, no. 1, p. 496, 2017.View at: Publisher Site | Google Scholar
L. Adane, S. Bhagat, M. Arfeen et al., “Design and synthesis of guanylthiourea derivatives as potential inhibitors of Plasmodium falciparum dihydrofolate reductase enzyme,” Bioorganic & Medicinal Chemistry Letters, vol. 24, no. 2, pp. 613–617, 2014.View at: Publisher Site | Google Scholar
L. Adane and P. V. Bharatam, “Modelling and informatics in the analysis of P. falciparum DHFR enzyme inhibitors,” Current Medicinal Chemistry, vol. 15, no. 16, pp. 1552–1569, 2008.View at: Publisher Site | Google Scholar
H. Park, H.-Y. Jung, S. Mah, and S. Hong, “Systematic computational design and identification of low Picomolar inhibitors of Aurora kinase a,” Journal of Chemical Information and Modeling, vol. 58, no. 3, pp. 700–709, 2018.View at: Publisher Site | Google Scholar
J. R. Schames, R. H. Henchman, J. S. Siegel, C. A. Sotriffer, H. Ni, and J. A. McCammon, “Discovery of a novel binding trench in HIV integrase,” Journal of Medicinal Chemistry, vol. 47, no. 8, pp. 1879–1881, 2004.View at: Publisher Site | Google Scholar
J. Vanichtanankul, S. Taweechai, J. Yuvaniyama et al., “Trypanosomal dihydrofolate reductase reveals natural antifolate resistance,” ACS Chemical Biology, vol. 6, no. 9, pp. 905–911, 2011.View at: Publisher Site | Google Scholar
J. Yuvaniyama, P. Chitnumsub, S. Kamchonwongpaisan et al., “Insights into antifolate resistance from malarial DHFR-TS structures,” Nature Structural & Molecular Biology, vol. 10, no. 5, pp. 357–365, 2003.View at: Publisher Site | Google Scholar
D. E. Hurt, J. Widom, and J. Clardy, “Structure of Plasmodium falciparum dihydroorotate dehydrogenase with a bound inhibitor,” Acta Crystallographica Section D Biological Crystallography, vol. 62, no. 3, pp. 312–323, 2006.View at: Publisher Site | Google Scholar
S. Forli, R. Huey, M. E. Pique, M. F. Sanner, D. S. Goodsell, and A. J. Olson, “Computational protein–ligand docking and virtual drug screening with the AutoDock suite,” Nature Protocols, vol. 11, no. 5, pp. 905–919, 2016.View at: Publisher Site | Google Scholar
L. S. Borquaye, E. N. Gasu, G. B. Ampomah et al., “Alkaloids from Cryptolepis sanguinolenta as potential inhibitors of SARS-CoV-2 viral proteins: an in silico study,” BioMed Research International, vol. 2020, Article ID 5324560, 14 pages, 2020.View at: Publisher Site | Google Scholar
O. Trott and A. J. Olson, “AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading,” Journal of Computational Chemistry, vol. 31, no. 2, pp. 455–461, 2010.View at: Publisher Site | Google Scholar
E. F. Pettersen, T. D. Goddard, C. C. Huang et al., “UCSF Chimera—a visualization system for exploratory research and analysis,” Journal of Computational Chemistry, vol. 25, no. 13, pp. 1605–1612, 2004.View at: Publisher Site | Google Scholar
A. L. Hopkins, C. R. Groom, and A. Alex, “Ligand efficiency: a useful metric for lead selection,” Drug Discovery Today, vol. 9, no. 10, pp. 430-431, 2004.View at: Publisher Site | Google Scholar
M. D. Shultz, “Setting expectations in molecular optimizations: strengths and limitations of commonly used composite parameters,” Bioorganic & Medicinal Chemistry Letters, vol. 23, no. 21, pp. 5980–5991, 2013.View at: Publisher Site | Google Scholar
C. Kutzner, S. Páll, M. Fechner, A. Esztermann, B. L. de Groot, and H. Grubmüller, “Best bang for your buck: GPU nodes for GROMACS biomolecular simulations,” Journal of Computational Chemistry, vol. 36, no. 26, pp. 1990–2008, 2015.View at: Publisher Site | Google Scholar
A. Amusengeri, R. B. Tata, and Ö. T. Bishop, “Understanding the pyrimethamine drug resistance mechanism via combined molecular dynamics and dynamic residue network analysis,” Molecules, vol. 25, no. 4, p. 904, 2020.View at: Publisher Site | Google Scholar
M. T. Che Omar, “Data analysis of molecular dynamics simulation trajectories of β-sitosterol, sonidegib and cholesterol in smoothened protein with the CHARMM36 force field,” Data in Brief, vol. 33, Article ID 106350, 2020.View at: Publisher Site | Google Scholar
R. Kumari, R. Kumar, and A. Lynn, “g_mmpbsa—a GROMACS tool for high-throughput MM-PBSA calculations,” Journal of Chemical Information and Modeling, vol. 54, no. 7, pp. 1951–1962, 2014.View at: Publisher Site | Google Scholar
W. Humphrey, A. Dalke, and K. Schulten, “VMD: visual molecular dynamics,” Journal of Molecular Graphics, vol. 14, no. 1, pp. 33–38, 1996.View at: Publisher Site | Google Scholar
P. J. Turner, XMGRACE, Version 5.1. 19, Center for Coastal and Land-Margin Research, Oregon Graduate Institute of Science and Technology, Beaverton, OR, USA, 2005.
E. Suresh Kumar, J. R. Etukala, and S. Y. Ablordeppey, “Indolo[3,2-b]quinolines: synthesis, biological evaluation and structure activity-relationships,” Mini Reviews in Medicinal Chemistry, vol. 8, no. 6, pp. 538–554, 2008.View at: Publisher Site | Google Scholar
A. L. Hopkins, “Network pharmacology: the next paradigm in drug discovery,” Nature Chemical Biology, vol. 4, no. 11, pp. 682–690, 2008.View at: Publisher Site | Google Scholar
N. T. Nguyen, T. N. H. Pham, N. T. Huy et al., “Autodock Vina adopts more accurate binding poses but Autodock4 forms better binding affinity,” Journal of Chemical Information and Modeling, vol. 60, no. 1, pp. 204–211, 2020.View at: Publisher Site | Google Scholar
E. Wang, H. Sun, J. Wang et al., “End-point binding free energy calculation with MM/PBSA and MM/GBSA: strategies and applications in drug design,” Chemical Reviews, vol. 119, no. 16, pp. 9478–9508, 2019.View at: Publisher Site | Google Scholar
G. Poli, C. Granchi, F. Rizzolio, and T. Tuccinardi, “Application of MM-PBSA methods in virtual screening,” Molecules, vol. 25, no. 8, p. 1971, 2020.View at: Publisher Site | Google Scholar
Y. Yuthavong, “Basis for antifolate action and resistance in malaria,” Microbes and Infection, vol. 4, no. 2, pp. 175–182, 2002.View at: Publisher Site | Google Scholar
S. Abbat, C. K. Jaladanki, and P. V. Bharatam, “Exploring PfDHFR reaction surface: a combined molecular dynamics and QM/MM analysis,” Journal of Molecular Graphics and Modelling, vol. 87, pp. 76–88, 2019.View at: Publisher Site | Google Scholar
G. B. Fogel, M. Cheung, E. Pittman, and D. Hecht, “Modeling the inhibition of quadruple mutant Plasmodium falciparum dihydrofolate reductase by pyrimethamine derivatives,” Journal of Computer-Aided Molecular Design, vol. 22, no. 1, pp. 29–38, 2008.View at: Publisher Site | Google Scholar
R. Ferreira de Freitas and M. Schapira, “A systematic analysis of atomic protein–ligand interactions in the PDB,” MedChemComm, vol. 8, no. 10, pp. 1970–1981, 2017.View at: Publisher Site | Google Scholar
S. Genheden and U. Ryde, “The MM/PBSA and MM/GBSA methods to estimate ligand-binding affinities,” Expert Opinion on Drug Discovery, vol. 10, no. 5, pp. 449–461, 2015.View at: Publisher Site | Google Scholar
J. M. J. Swanson, R. H. Henchman, and J. A. McCammon, “Revisiting free energy calculations: a theoretical connection to MM/PBSA and direct calculation of the association free energy,” Biophysical Journal, vol. 86, no. 1, pp. 67–74, 2004.View at: Publisher Site | Google Scholar
T. Lemcke, I. T. Christensen, and F. S. Jørgensen, “Towards an understanding of drug resistance in malaria: three-dimensional structure of Plasmodium falciparum dihydrofolate reductase by homology building,” Bioorganic & Medicinal Chemistry, vol. 7, no. 6, pp. 1003–1011, 1999.View at: Publisher Site | Google Scholar
M. D. Parenti, S. Pacchioni, A. M. Ferrari, and G. Rastelli, “Three-dimensional quantitative structure- activity relationship analysis of a set of Plasmodium falciparum dihydrofolate reductase inhibitors using a pharmacophore generation approach,” Journal of Medicinal Chemistry, vol. 47, no. 17, pp. 4258–4267, 2004.View at: Publisher Site | Google Scholar
G. Rastelli, W. Sirawaraporn, P. Sompornpisut et al., “Interaction of pyrimethamine, cycloguanil, WR99210 and their analogues with Plasmodium falciparum dihydrofolate reductase: structural basis of antifolate resistance,” Bioorganic & Medicinal Chemistry, vol. 8, no. 5, pp. 1117–1128, 2000.View at: Publisher Site | Google Scholar
E. Bilsland, L. van Vliet, K. Williams et al., “Plasmodium dihydrofolate reductase is a second enzyme target for the antimalarial action of triclosan,” Scientific Reports, vol. 8, no. 1, pp. 1–8, 2018.View at: Publisher Site | Google Scholar
M. Hoarau, J. Vanichtanankul, N. Srimongkolpithak, D. Vitsupakorn, Y. Yuthavong, and S. Kamchonwongpaisan, “Discovery of new non-pyrimidine scaffolds as Plasmodium falciparum DHFR inhibitors by fragment-based screening,” Journal of Enzyme Inhibition and Medicinal Chemistry, vol. 36, no. 1, pp. 198–206, 2021.View at: Publisher Site | Google Scholar
R. Baumgartner, M. Walloschek, M. Kralik et al., “Dual binding mode of a novel series of DHODH inhibitors,” Journal of Medicinal Chemistry, vol. 49, no. 4, pp. 1239–1247, 2006.View at: Publisher Site | Google Scholar
P. Norman, “Evaluation of WO2013076170: the use of a dihydroorotate dehydrogenase inhibitor for the treatment of psoriasis,” Expert Opinion on Therapeutic Patents, vol. 23, no. 10, pp. 1391–1394, 2013.View at: Publisher Site | Google Scholar
Y. Zhou, L. Tao, X. Zhou et al., “DHODH and cancer: promising prospects to be explored,” Cancer & Metabolism, vol. 9, no. 1, pp. 22–25, 2021.View at: Publisher Site | Google Scholar