Exploring anti-malarial potential of FDA approved drugs: an in silico approach

Dataset

Information on gene products encoded in the genome of P. falciparum 3D7 isolate was obtained from PlasmoDB (release 24) [10]. Proteins annotated as pseudogenes were excluded as such entries correspond to genes that apparently lack protein-coding ability and are unlikely to be expressed in the cell. The protein sequences corresponding to 5400 protein-coding genes were thus retrieved from the database. Furthermore, protein expression profiles of the parasite proteins during their developmental stages in human were extracted from the published works of Pease et al. [11], Florens et al. [12] and Le Roch et al. [13] for intra-erythrocytic stages (ring, trophozoite, schizont, merozoite), from Lindner et al. [14] for liver stage (sporozoite), and from Silvestrini et al. [15] for the final gametocyte stage. Details on FDA-approved drugs (1634) and the associated protein targets (1573) were obtained from DrugBank (version 4.3), an extensive resource on drug and target information.

Approach to recognize potential drug-target interactions

The protocol adopted to identify FDA-approved drugs that can be repurposed against proteins in P. falciparum is similar to the protocol employed in an earlier published report on repurposing drugs against Mycobacterium tuberculosis H37Rv [16]. The exclusion of drugs known to be effective against human proteins and the exclusion of anti-malarial agents currently in use (Table 1) formed the initial step of the approach. Also, only those drugs that hold evidence on pharmacological validation against established targets were considered. Thus, 196 FDA-approved drugs associated with 138 protein targets were considered for subsequent analysis. The likelihood of drugs to be active as potential anti-malarial agents is dependent on the similarity of known targets to P. falciparum proteins. To recognize target-related proteins in P. falciparum, an iterative sequence search program, jackhammer availed through HMMER3.0 suite [17], was employed at an E-value cut-off of 0.0001 and five rounds of iteration. An alignment coverage cut-off of 70% or an alignment encompassing at least one functional and/or structural domain was used as an additional criterion to eliminate unreliable hits characterized by short stretches of alignment.

Once reliable target-related proteins are recognized in P. falciparum, a comparative structural analysis was undertaken to pinpoint probable drug-binding sites in the parasite proteins, based on information on ligand-binding sites in the known targets. Structural information for target proteins and parasite proteins was obtained from PDB (Protein Data Bank). For parasite proteins of no known structure, structural models were retrieved from ModBase [18] or built using MODELLER v.9.14 [19]. The structural models obtained were checked for reliability based on z-DOPE (Discrete Optimized Protein Energy) score [20] (<0), ModPipe Quality Score (MPQS) cut-off of 1.1 and query coverage threshold of 80% or a query coverage of at least one functional and/or structural domain. Comparative evaluation of binding sites across established targets and their homologues in P. falciparum was pursued depending on the availability of crystal structures of ligand-bound targets. A structural alignment algorithm, TM-align [21], was employed to assess the extent of conservation of residues in local aligned regions between the parasite proteins and corresponding related targets. A TM-score cut-off of 0.5 was used, which typically implies convincing structural similarity of the proteins aligned [22]. In certain cases, a reliable structural model for the P. falciparum protein or its domain could not be built, primarily due to abundance of low-complexity regions in the protein and/or presence of non-conserved inserts within the functional domains. Indeed, these features contribute to the pronounced sequence divergence observed in genome of P. falciparum [23–25]. Thus, in such cases, the aligned regions of interest between the protein sequences of P. falciparum and related targets were manually examined to infer feasibility of drug-target interactions in P. falciparum. Similarly, in the absence of crystal structures of established targets, potential drug-target interactions in the parasite proteins were conjectured on the basis of alignment of functionally relevant regions.

Potential repurpose-able drug candidates

A total of 71 FDA-approved drugs, predominantly constituted by antibacterial agents (56), were recognized as repurpose-able candidates against 89 P. falciparum proteins. The complete list of predicted drug-target associations is provided in Additional file 1. At least five of the anti-malarial drugs listed in Table 1 are anti-bacterial and have been repurposed for use against malaria. Based on hierarchical, rule-based, structural classification of drugs, as obtained from DrugBank [26], the 71 drugs could be grouped into 22 drug classes. Figure 1 outlines the distribution of 71 drugs under the 22 classes, along with number of predicted parasite targets associated with each drug class. The predominance of antibacterial agents, represented as green coloured bars, can be readily inferred from the Figure. Notably, a number of instances can be observed in the Figure, which corresponds to drugs with polypharmacological potential, i.e., the ability to act on multiple targets. The most striking example in the Figure is that of two antibacterial drugs (DB00410: mupirocin, DB00548: azelaic acid) under the class ‘Fatty acyls’ which have been predicted to target 21 parasite proteins; of which nine parasite proteins involved in aminoacylation are potential targets for mupirocin and 12 metabolic proteins of the parasite are potential targets for azelaic acid.

Mupirocin, an antibiotic isolated from Pseudomonas fluorescens, is one of the widely used bacteriocidal agent known to inhibit isoleucyl-tRNA synthetase (also known as isoleucyl-tRNA ligase). Recently, a study demonstrated the potent inhibitory activity of mupirocin against blood-stage P. falciparum [27]. Despite its effectiveness, mupirocin is unfit to serve as a lead for anti-malarial therapy as it is restricted to topical usage due to rapid hydrolysis in body fluids. Nevertheless, non-hydrolysable mupirocin analogues with evidence on inhibitory activity can be explored for their potential use against P. falciparum. Indeed, two analogues of mupirocin, CHEMBL38412 and CHEMBL315230, were identified from an earlier published study [16] where two-step in silico screening (two-dimensional followed by three-dimensional molecular similarity calculations) was employed on a set of FDA-approved drugs against a database of ChEMBL compounds [28, 29], which hosts exhaustive details on drug-like bioactive compounds largely extracted from literature and other databases. Of the two mupirocin analogues CHEMBL38412 and CHEMBL315230, which are experimentally verified inhibitors of isoleucyl-tRNA synthetase [30, 31], CHEMBL38412 is reportedly a non-hydrolysable analogue of mupirocin and exhibits similar antibacterial activity in vitro as mupirocin [30]. Elucidations on binding site and binding pose of CHEMBL38412 can be readily inferred based on structural similarity between mupirocin-bound isoleucyl-tRNA synthetase and related parasite proteins (Table 2 and Additional file 2: Figure S1) and the extent of molecular similarity between the drug and its analogue as illustrated in Fig. 2. Based on reasonable structural similarity amongst related proteins as well as the compounds, the mupirocin analogue CHEMBL38412 forms an attractive lead and thus can be investigated further for its anti-malarial activity.

Several other drug classes corresponding to drugs with possible polypharmacological potential, as delineated in Fig. 1, can be investigated further. Figure 1 also illustrates populated drug classes including ‘Quinolines’, ‘Carbohydrate conjugates’, ‘Benzene derivatives’ and ‘Imidazopyrimidines’ with relatively fewer predicted targets associated with them. Indeed, this observation implies the possibility of combination therapy. But the risk of cross-resistance exists, where resistance to one drug may confer resistance to other chemically similar drugs or those drugs that share similar mechanism of action. Reports on tetracycline and doxycycline resistance, alone or in combination with other drugs, limits the possibility of using approved drugs chemically similar to tetracycline. Seven such drugs (DB00595: oxytetracycline, DB01301: rolitetracycline, DB00618: demeclocycline, DB00256: lymecycline, DB00453: clomocycline, DB01017: minocycline, DB00560: tigecycline) in the drug class ‘Tetracyclines’ were recognized in the current analysis as potential drug candidates against ten potential parasite targets (Fig. 1). Thorough experimental investigations are necessary to ascertain the likelihood of these drugs as repurpose-able candidates.

Drug targets recognized in Plasmodium falciparum

The 89 potential targets identified in P. falciparum could be categorized under six functional classes, including protein biosynthesis, metabolic processes, DNA-dependent activities, microtubule-associated processes, transport and proteolytic activities, most of which are expressed in all six life-cycle stages of infection in human host. Figure 3 illustrates the distribution of predicted targets across the six life-cycle stages of P. falciparum binned under six functional categories. As shown in Fig. 3, 38 of 89 predicted targets participate in cellular protein synthesis, which are predominantly expressed in all six life-cycle stages of the parasite. This category includes proteins participating in aminoacylation and protein translation, essential for housekeeping activities. Other proteins of parasite housekeeping machinery including DNA polymerases, DNA topoisomerases, DNA gyrases, microtubules and structural constituents of cytoskeleton, qualify as attractive targets as these P. falciparum proteins expressed in all six life-cycle stages during infection in humans play a significant role in regulating rate of translation, cell growth and cell development [12, 23]. Thus, targeting such parasite proteins (64) belonging to the functional classes ‘Protein biosynthesis’, ‘DNA-dependent processes’ and ‘Microtubule-associated’ (Fig. 3) can impede growth and transition between different developmental stages of the parasite. Based on the current analysis, a list of 46 drugs were identified, most of which are antibacterial agents, predicted to target 64 proteins involved in parasite housekeeping machinery (Additional file 1).

An important characteristic of P. falciparum is its ability to undergo vast re-organization of metabolic make-up during the course of its life cycle in multiple host environments. Inhibition of critically important metabolic enzymes can be detrimental for the parasite’s survival, one of which is the glycolytic enzyme lactate dehydrogenase (pfLDH) essential for intra-erythrocytic growth of the parasite [32], recognized as a target for the drug nitrofural (DB00336) in the current analysis. Studies on pfLDH have reported a distinct nature of NADH binding pocket [32] where the anti-malarial chloroquine is believed to bind competitively [33]. Nitrofural, a potent antibacterial agent, is known to inhibit several bacterial enzymes involved in aerobic and anaerobic degradation of glucose and pyruvate. Crystal structure of nitrofural-bound nitroreductase enzyme in Escherichia coli demonstrated the accommodation of the drug in the cofactor binding site [34]. Thus, the cofactor binding site in the crystal structure of pfLDH, i.e., the NADH binding pocket, along with adjacent solvent exposed cavities, were used as receptor grids to dock nitrofural using Glide (Grid-based ligand docking with energetics) [35–37] and assess the predicted binding energy and binding pose of the approved drug. Using Glide XP (extra precision) scoring function [37], the predicted binding site of the drug nitrofural was evaluated. Figure 4 illustrates the docked pose of nitrofural, achieved with a best possible energy score of −6.97 kcal/mol, at the NADH binding pocket of pfLDH. The docked drug was identified to make contacts with critical residues of pfLDH that are involved in cofactor binding, including Asp53 and Thr97 (Fig. 4c, d), implying the likeliness of nitrofural in inhibiting the glycolytic enzyme of the parasite. In addition to pfLDH, six P. falciparum metabolic proteins, including malate dehydrogenase (PF3D7_0618500), lactate dehydrogenase (PF3D7_132520), lipoamide dehydrogenase (PF3D7_0815900), thioredoxin reductase (PF3D7_0923800), dihydrolipoyl dehydrogenase (PF3D7_1232200), and glutathione reductase (PF3D7_1419800), several of which are essential for the parasite’s growth during intra-erythrocytic development stages, were recognized as potential targets for nitrofural. Considering such a polypharmacological potential of nitrofural, further experimental investigations on this FDA-approved drug can reveal insights into its usefulness as an anti-malarial agent. Other metabolic enzymes identified as targets include enoyl-acyl carrier reductase, dihydropteroate synthetase, thioredoxin reductase, bifunctional dihydrofolate reductase–thymidylate synthase, and other oxido-reductive enzymes which have been regarded as attractive targets for anti-malarial drug development [38], many of which are expressed in all six stages of the parasite’s life cycle (Fig. 3). The identification of such pharmaceutically relevant metabolic proteins justifies the strength of the approach used and thus, the feasibility of approved drugs predicted to target such proteins can be further investigated.

Plasmodium falciparum comprises a relict-plastid, termed as the apicoplast, which is indispensable for the parasite’s growth and typically contains a range of metabolic pathways and housekeeping processes that differ from those in human hosts [39]. Thus, parasite proteins pertaining to the apicoplast present ideal strategies for drug design. Of 89 predicted targets, 29 were identified as constituents of apicoplast—8 metabolic enzymes and 21 involved in housekeeping machinery. These entries are highlighted in Additional file 1. Information on subcellular localization of other parasite proteins was obtained from Malaria Metabolic Pathways database [40]. Figure 5 illustrates the distribution of number of predicted targets in terms of their subcellular localization in P. falciparum (inner ring), along with information on upregulated, intra-erythrocytic stage-specific expression of associated proteins (outer ring). As shown in Fig. 5, more than half of the predicted P. falciparum targets are localized in subcellular organelle: nucleus, apicoplast and mitochondrion.

The predictions were compared with previously published studies on recognizing anti-malarial potential of approved drugs [8, 43]. Nine drug-target associations were identified pertaining to apicoplast proteins and anti-tubercular drugs (rifamixin, rifampicin) concurring with the predictions made in the current study which, in addition to the set of drugs mentioned in Table 3, can be readily taken through further stages of anti-malarial drug development pipeline.