Structural basis of malaria parasite phenylalanine tRNA-synthetase inhibition by bicyclic azetidines

The inhibition of Plasmodium cytosolic phenylalanine tRNA-synthetase (cFRS) by a novel series of bicyclic azetidines has shown the potential to prevent malaria transmission, provide prophylaxis, and offer single-dose cure in animal models of malaria. To date, however, the molecular basis of Plasmodium cFRS inhibition by bicyclic azetidines has remained unknown. Here, we present structural and biochemical evidence that bicyclic azetidines are competitive inhibitors of L-Phe, one of three substrates required for the cFRS-catalyzed aminoacylation reaction that underpins protein synthesis in the parasite. Critically, our co-crystal structure of a PvcFRS-BRD1389 complex shows that the bicyclic azetidine ligand binds to two distinct sub-sites within the PvcFRS catalytic site. The ligand occupies the L-Phe site along with an auxiliary cavity and traverses past the ATP binding site. Given that BRD1389 recognition residues are conserved amongst apicomplexan FRSs, this work lays a structural framework for the development of drugs against both Plasmodiumand related apicomplexans.

Malaria, caused by apicomplexan parasites of the genus Plasmodium, presents a formidable global health challenge mainly due to the emergence of parasite strains that are resistant to front-line drugs1,2. It is therefore necessary to discover and validate new drug targets as well as compounds whose efficacy is unaffected by mechanisms of resistance to traditional antimalarials2,3. Ideally, an antimalarial development candidate should have a new mechanism of action (MoA) with rapid asexual blood-stage parasite reduction, and activity against all stages of the parasite lifecycle in the human host2. Recently, a series of small molecules based on a bicyclic azetidine core have been discovered that exhibit multistage antimalarial activity and can achieve single-dose cures in a mouse model of malaria2. It was demonstrated that these compounds exert their antimalarial activity via inhibition of Plasmodium falciparum cytosolic phenylalanyl-tRNA synthetase (cFRS)2, an enzyme essential for protein synthesis. The PfcFRS was indeed validated both genetically and biochemically as a drug target for the bicyclic azetidine series of molecules2, setting the stage for further drug discovery efforts. Aminoacyl-tRNA synthetases (aaRSs) activate amino acids as aminoacyl adenylates (AA-AMP), and enable their relay to the 3′-ends of cognate tRNAs as feed for ribosomes3,4. Inhibition of aaRSs therefore results in the interruption of cell growth and ultimately in cell death3. Of note, Plasmodium aaRSs other than FRS have also recently become the focus of antimalarial development efforts3.

As previously shown for Plasmodiam falciparum, the parasite has three protein translation compartments: in the cytoplasm, apicoplast and in the mitochondria where FRSs reside to feed charged tRNAs into ribosomal-based protein synthesis4. In both P. falciparum and P. vivax, FRSs exist as heterodimers of alpha (α) and beta (β) subunits that further dimerize to form a complex of (αβ)24. This hetero-tetrameric organization of cFRSs is conserved but with significant differences in the chain lengths and functions of α and β subunits5. The FRS α subunit contains the active site and catalyzes the two-step aminoacylation reaction, while the main functions of the β subunit are to recognize the anticodon region of tRNA and to edit mischarged tRNA molecules with isosteric amino acids such as tyrosine5. The cFRSs are highly conserved and exhibit high sequence identity amongst Plasmodium species suggesting that FRS from all five parasites causative of human malaria can be targeted by a single chemical series. In this work, we reveal the biochemical and structural basis of inhibition of Plasmodium cFRS via the x-ray co-crystal structure of PvcFRS with BRD1389, a potent antimalarial from the bicyclic azetidine chemical series.

BRD1389 binds cFRS selectively and inhibits aminoacylation via L-Phe competition

To determine the mode of inhibition of cFRS by bicyclic azetidines, we turned to BRD1389 (Fig. 1a, Supplementary Fig. 1, 2), a series analogue with high in vitro potency in the growth inhibition assay (Pf3D7 and PfK1 EC50 are 13 and 12 nM respectively), potent abrogation of parasite FRS enzyme activity, and high selectivity index over the human orthologue (Table 1, Fig. 1b, Supplementary Fig. 3c, d). We measured the enzymatic activity of purified heterodimeric PfcFRS using malachite green-based aminoacylation assays with substrates L-Phe and ATP6. Specifically, in order to understand PfcFRS kinetics, enzyme activity was evaluated at a fixed saturating concentration of one substrate (L-Phe or ATP) and varying concentrations of the other at different inhibitor concentrations. The experimental data were then assessed for modes of competitive, mixed, or uncompetitive inhibition. We first performed this study at a constant concentration of ATP with increasing concentrations of L-Phe. These data indicate that BRD1389 displays a mode of competitive inhibition vis-a-vis L-Phe binding (Fig. 1c) as fitting to a global competitive inhibition model resulted in Ki of 6 ± 2 nM for the compound. These results are consistent with previously reported data on in vitro parasite growth inhibition by a closely related bicyclic azetidine (BRD1095) which showed higher EC50 values with increasing concentrations of L-Phe2. We then repeated the mode of inhibition study using ATP as the variable-concentration substrate. The data were then fit to the modified high-substrate inhibition Michaelis–Menten equation (for the corresponding double-reciprocal plot, see Fig. 1d), consistent with a non-competitive inhibition model, where BRD1389 gave Ki value of 10 ± 5 nM. Taken together, these results suggest that BRD1389 preferentially binds to the free enzyme via competition of L-Phe. This mechanism of action is unique when compared to other promising antimalarial aaRS inhibitors of cladosporin class7,8,9 (it mimics adenosine) or the halofuginone scaffold (requires ATP for tight binding)10,11,12. Additionally, in order to understand the degree of selectivity, binding kinetics were measured using PfcFRS and human cytosolic FRS (HscFRS). The data exhibits quite distinct binding affinities as well as binding kinetics (BRD1389 Kd: PfcFRS: 4 nM; HscFRS: 16 µM) when fit to a 1:1 binding model of nonlinear regression (specific binding) (Supplementary Fig. 3c, d).

Overall structure of PvcFRS

Toward unravelling the structural basis of the protein–inhibitor interaction, we crystallized the PvcFRS-BRD1389 complex and obtained crystals that diffracted to 3 Å. The structure was solved by molecular replacement (MR) using HscFRS as a template (PDB: “3L4G [https://www.rcsb.org/structure/3l4g]”). The PvcFRS-BRD1389 crystals belong to orthorhombic space group P21212 with one heterodimer (αβ) per asymmetric unit. The (αβ)2biological heterotetrametric assembly is completed via the crystallographic two-fold axis along c. The PvcFRS (αβ)2 assembly is consistent with the size exclusion chromatography profile of purified protein, where it elutes at a size of ~298 kDa (Supplementary Fig. 3a). Our final PvcFRS-BRD1389 atomic model contains 905 residues- 299:α subunit, 606:β subunit, one Mg2+, and one ligand (BRD1389). The N-terminal domain (residues 1–270) of the α subunit was absent in the crystal and no electron density was observed for it. This was further verified by crystal packing analysis of PvFRS-BRD1389 complex and via SDS-PAGE of crystallised protein (Supplementary Fig. 4). The overall fold and organization of α and β subunit is very similar to that of the human orthologue (HscFRS, Fig. 2a, b) with the exception that the PvcFRS β subunit lacks an 18-residue-long fragment within its PB1 domain whereas it has two unique insertions (IL-1: 421–444, IL2: 535–551) in its B2 sub-domain (Supplementary Fig. 5). Intriguingly, the association of β1 and β2 subdomains (of the β subunit) with α subunit is significantly different between the parasite and human enzymes (Fig. 1e–j, Supplementary Fig. 6a). In HscFRS, the α subunit is enclosed by β1 and β2 subdomains of the same β subunit, whereas in PvcFRS β1 and β2′ of the β subunit is associated with the α subunit (Fig. 1j, i). This difference arises due to the shorter linker length between β1 and β2 of PvcFRS when compared to the HscFRS (Fig. 1h, k) which has a three residue (384-TYT-386) insertion. Interestingly, this three-residue shorter linker is observed in all human malaria parasites (Fig. 1h, k) and is suggestive of domain swapping (Fig. 1g) in PvcFRS to form a closed, functional hetero-tetrameric (αβ)2 assembly.

BRD1389 occupies the L-Phe site and an auxiliary site on PlasmodiumcFRS

During the refinement of PvcFRS model, an investigation of difference Fourier electron density (Fo-Fc) maps revealed BRD1389 to be bound at the enzyme active site (Fig. 2c, d). The bound ligand was subsequently verified by simulated annealing omit (SA-omit) map (Fig. 2d). Superposition of the PvcFRS α subunit on HscFRS and Thermus thermophilus FRS (TtFRS, PDB: “2IY5 [https://www.rcsb.org/structure/2IY5]”) allowed the mapping of L-Phe, ATP and tRNA binding sites on these FRSs (Fig. 3c). Further analysis of the PvcFRS:BRD1389 complex revealed that the diarylacetylene moiety of BRD1389 occupies the L-Phe site, its 4-cyclopropoxy phenyl resides in an auxiliary pocket, and the [6.2.0]-diazabicyclodecane group skirts the ATP site in PvcFRS (Fig. 3d, Supplementary Fig. 7). The PvcFRS:BRD1389 complex is stabilized mainly by hydrophobic and hydrophilic interactions at multiple sites that contribute towards recognition of the diazabicyclodecane core, the 4-cyclopropoxyphenyl and the diarylacetylene appendages (Fig. 3a).

The active site of BRD1389-bound Plasmodium cFRS adopts unique conformations

In this complex, open and close conformation for residue Arg548 (Fig. 4a, d) is noticeable. In particular, the flexing of Arg548 likely opens the entry point of auxiliary pocket for 4-cyclopropoxyphenyl accommodation, underscoring the induced-fit nature of BRD1389 interaction with PvcFRS (Fig. 4d). In phenyladenylate-bound TtFRS, both the guanidinium moiety of Arg321 and Phe216 (PvcFRS; Phe455) provide stacking support to the adenine ring of ATP, whereas in PvcFRS-BRD1389 the corresponding guanidinium (belonging to Arg548) is displaced away from the active site adopting an open conformation (Fig. 4a, d). This can be further investigated in L-Phe-bound HscFRS, where the corresponding Arg463 moves inward (i.e., towards the active site) adopting instead a closed conformation. Additionally, in the PvcFRS:BRD1389 complex there are two noticeable, ordered loop conformations in proximity to BRD1389 binding site within the PA1 domain: (1) a left-hand outward displacement (open conformation) of residues 443–453 (loop 1, in ATP binding pocket), and (2) a left-hand inward movement (closed) of residues 507–515 (loop 2, in auxiliary pocket) (Fig. 4a–c, Supplementary Fig. 6b, c). In PvcFRS:BRD1389, loop 2 adopts a closed conformation akin to phenyladenylate-bound TtFRS and unlike L-Phe-bound HscFRS (open conformation, Fig. 4c, Supplementary Fig. 6c), whereas loop 1 moves relative to both TtFRS and HscFRS to achieve an open state (Fig. 4b, Supplementary Fig. 6b).

Structural basis of BRD1389 recognition by PvcFRS

The diarylacetylene moiety of BRD1389 that occupies the L-Phe site is recognized via a bed of β-strand residues Ala541-Trp542-Gly543-Leu544 (Fig. 3b), while its edge phenyl ring is covered by Tyr497 where it provides an aromatic T-shaped π…π interaction at a distance of ~4.2 Å (Fig. 3b). In addition, the diarylacetylene moiety is surrounded by hydrophobic elements of Asn519, Gln457, and Glu459 (Fig. 3b). The urea moiety of BRD1389 is buried in a groove that is composed of Gly506, His508, Glu510, Lys512–513, Leu515, Ile552 and Pro549 (Figs. 3b, e and 4c). The urea/groove interactions are mainly stabilized by hydrophobic contacts, particularly from Pro549 and His508 while residues Val517, Ile483 and Pro549 provide sandwich support to the cyclopropoxy moiety (Figs. 3b and 4c). The urea component of BRD1389 forms hydrogen bonding interactions with main-chain N-atom of Ser545 (Fig. 3a, b). The above sets of extensive interactions position BRD1389 in an ‘L’ shaped conformation wherein its methoxy methyl group is surrounded by socket residues Arg443, Glu445, His451, and Phe455. Interestingly, the crystallographic pose of BRD1389 is close to the conformation that the inhibitor is predicted to adopt in aqueous solution (Supplementary Fig. 8 and Supplementary Table 2). This similarity highlights the importance of the three-dimensional shape and rigidity of the diazabicyclodecane scaffold in pre-orienting the molecular appendages for optimal target engagement. From overlaying the structure of PvcFRS:BRD1389 with that of phenyladenylate-bound TtFRS, it is apparent that the diazabicyclodecane core and its methoxymethyl extension partially brush past the adenine binding region of the canonical ATP binding site (Figs. 3d and 4b). Given the high binding affinity of BRD1389 for PvcFRS (4 nM, Table 1), it is feasible that BRD1389 blocks the interaction of Plasmodium cFRS with L-Phe first and then with ATP. Indeed, upon incubation of PvcFRS with high concentrations of both BRD1389 and an ATP analogue (the non-hydrolysable adenosine 5’-(β,γ-imido) triphosphate, i.e. AMPPNP) we observed only binding of BRD1389. This result further supports that BRD1389 binding may occlude ATP engagement. Strikingly, residues in PvcFRS that recognize key ligand components (diazabicyclodecane core, 4-cyclopropoxy phenyl, methoxymethyl and diarylacetylene moieties) are conserved across the apicomplexan phyla, including in human-infecting parasites such as Toxoplasma and Cryptosporidium (Supplementary Fig. 9).

Basis of selectivity and resistance-conferring mutations

Next, to understand the structural basis of selective binding and inhibition by bicyclic azetidines of Plasmodium cFRS versus the human orthologue, we compared atomic structures of PvcFRS:BRD1389 and HscFRS-L-Phe complexes focusing on residues located within 5 Å of the ligand site. Three variant residues Pv-V458/Hs-I373, Pv-Y480/Hs-F395, Pv-I483/Hs-L398 are located within an auxiliary pocket of PvcFRS (Fig. 5a). This observation indicates that the selectivity of BRD1389 may arise from the terminal cyclopropyl ether (Fig. 5a). Significantly, all protein residues known to confer resistance to bicyclic azetidines upon mutation2 are located within the α subunit of FRS, in proximity to the BRD1389 (Fig. 5b). None of these mutations are directly in the ATP, tRNA, or L-Phe sites (Fig. 5b, Supplementary Table 3). However, one significant mutation PvcFRS-L544V (equivalent to PfcFRS-L550V) that diminishes BRD1389 potency structurally underpins the [6.2.0]-diazabicyclodecane ring of BRD1389 (Table 1, Fig. 5b, d). When we reconstructed the PfcFRS-L550V and tested its enzymatic fidelity, we observed a significant increase in Km value (10-fold) for the substrate L-Phe, but not for ATP, indicating that resistance to bicyclic azetidines may arise from a trade-off between parasite survival and the essential nature of cFRS enzymatic activity (Fig. 5c, d).

a Close-up view of PvcFRS (cornflower blue) with HscFRS (grey, PDB “3L4G [https://www.rcsb.org/structure/3l4g]”) as superimposed structures. Non-conserved residues within 5 Å of BRD1389 are shown. b PvcFRS structure with known PfcFRS mutations are highlighted in red. All resistance mutations map within the confines of the active site region wherein the L550V mutant lies closest to the bound ligand. c, d Effect of L550V mutation on enzyme activity. Both enzymes show almost similar Km value for the ATP (18 ± 1 µM for PfcFRS wild type (WT) and 11 ± 1 µM for PfcfRS mutant (L550V)) whereas the Km value for L-Phe is increased ~10-fold for the mutant enzyme (0.25 ± 0.017 µM for PfcFRS (WT) and is 1.85 ± 0.84 µM for PfcFRS mutant (L550V)). Data are shown as mean ± SD (n = 3 independent experiments). Source data are provided as a Source Data file.

Full size image

Through a combination of biochemical and crystallographic studies, we have revealed the molecular underpinning of Plasmodium cFRS inhibition by bicyclic azetidines. We have shown that BRD1389 inhibits parasite cFRS function by primarily blocking the binding of L-Phe in a competitive manner. The diphenylacetylene moiety of BRD1389 occupies the L-Phe binding site while the [6.2.0]-diazabicyclodecane core partially occludes the ATP binding region. The cyclopropoxyphenyl urea region of BRD1389, in turn, occupies an auxiliary pocket in PvcFRS. Residue variations between the malaria parasite cFRS and the human orthologue in this region underpin the highly selective enzyme inhibition and parasite killing by bicyclic azetidines. Two classes of malaria parasite aaRS inhibitors have been structurally evaluated to date3. These act either as single-site occupants (cladosporin, an adenosine mimic) or dual site engagers (halofuginone, a mimic of L-Pro and 3′ end of tRNA). As BRD1389 occupies both the L-Phe site and an auxiliary pocket within PvcFRS, it represents a novel dual-site malaria parasite aaRS inhibitor (Fig. 6a, 6; right panel).

Overall, our mapping of protein regions and residues that contribute both to cFRS inhibitor selectivity and drug resistance provides a structural platform for designing the next generation of compounds with improved potency and safety profiles. Indeed, the enzyme-inhibitor structure also reveals how the underlying principles of the diversity-oriented synthesis (DOS)13,14,15 library that yielded this inhibitor scaffold, i.e., inclusion of rigid bicyclic skeletons and multiple stereogenic elements—played a key role in accessing pockets within the enzyme that may have been inaccessible by compounds in classical libraries. Traditional libraries are replete with compounds that have a high percentage of atoms with sp2 hybridization, leading to flatter architectures. BRD1389, which results from DOS, in contrast makes sharp turns and penetrates into deep pockets within PvcFRS that are nearly at right angles.

This work may thus allow generation of compound libraries with tuneable drug-like properties that can focus on other apicomplexan-driven human diseases via targeting their FRSs. In particular, guided by structure, triple site inhibitors can now be developed that fully occupy the ATP site via chemical modifications of the [6.2.0]-diazabicyclodecane scaffold. More generally, this work paves the way for novel drug development against malaria and, potentially, other diseases caused by apicomplexans such as toxoplasmosis and cryptosporidiosis.

Prediction of BRD1389 conformation

Calculations were performed using Schrödinger Maestro Version 12.3.012, MMshare Version 4.9.012, Release 2020-1, Platform Windows-x64. Geometry optimization was performed in Jaguar16 Version 10.7, Release 12, at ultrafine accuracy level, using density functional theory (B3LYP), the 6–31 G** basis set, and the Poisson–Boltzmann Finite (PBF) water solvation model. The crystallographic and computed structures were superposed by Maximum Common Substructure using the Schrödinger Maestro Superposition tool.

Synthesis of BRD1389

BRD1389 was prepared from known compound (8 R,9 R,10 S,Z)-9-(4-bromophenyl)-6-((4-nitrophenyl) sulfonyl)-10-((trityloxy)methyl)-1,6-diazabicyclo[6.2.0]dec-3-ene.14 Please refer to the Supplementary Information for synthetic protocols and analytical data.

Protein expression and purification

Full-length PfcFRS was purified according to the earlier published report2. Full-length PvcFRS was also purified according to the same protocol. In brief, the genes encoding PvcFRS alpha subunit (PVX_081300) and beta subunit (PVX_090880) were cloned into E. colipETM11 and pETM20 plasmids respectively. Both plasmids were co-transformed into E. colistrain B834 and were induced overnight for overexpression with 0.5 mM isopropyl-β-D-thiogalactoside (IPTG) at 16 ° C for 18 h. The E. coli cell lysate was first loaded onto a nickel–nitrilotriacetic (Ni–NTA) column (GE Healthcare) and the eluted fraction was further purified with Heparin chromatography (GE Healthcare) to a single band as indicated by SDS–polyacrylamide gel electrophoresis with Coomassie brilliant blue staining. The purified protein was found as a single peak with the elution volume consistent of a homogeneous PvcFRS hetero-tetramer on the Superdex 200 analytical gel filtration column (GE Healthcare). The purified PvcFRS was concentrated to 25 mg ml−1 and stored at −80 °C in 25 mM HEPES buffer, pH 7.5, 200 mM NaCl, 5 mM βME.

Evaluation of parasite growth inhibition

Plasmodium falciparum strain 3D7 (chloroquine–sensitive) and K1 (chloroquine-resistant) were obtained from Kitasato University and used for testing antimalarial activities in vitro. The cultivation of P. falciparum was conducted according to Trager’s method with some modification17. Precisely, parasites were kept in culture flasks with RPMI1640 medium supplemented with 10% human plasma and 2% fresh human erythrocytes and incubated at 37 °C with the gas condition of 5% CO2 and 5% O2. The parasitemia (percentage of infected erythrocytes to total erythrocytes) were kept within 0.25–10%. Culture medium was replaced, and fresh erythrocytes were supplied every 2–3 days. Drug susceptibility test was conducted according to Desjardins’s method18 with some modification. The bicyclic azetidines and known antimalarial agents (chloroquine and artemisinin) were tested at the same time. Precisely, 199.5 μl of parasite cultures (2% hematocrit and 0.75–1% parasitemia) and 0.5 µl of compound solution serially diluted in DMSO were poured into every well in 96-well titer plates and final drug concentrations were set within 0.001–1 μg ml−1. The plates were kept at 37 °C with the gas condition of 5% CO2 and 5% O2 for 72 h and then parasite growth was quantified with Makler’s method to detect plasmdoial lactate dehydrogenase19with some modification. Precisely, culture plates were kept in freezer overnight and then thawed at 37 °C to disrupt the erythrocytes and parasite cells. In the new 96-well titer plates, 100 μl of enzyme reaction solution (110 mM Li-lactate, 0.5 mM acetylpyridine-adenine dinucleotide, 50 mM Tris (pH 7.5), 10 mM EDTA, 50 mM KCl, and 15 g l−1 PEG6000) and 20 µl of freeze-thaw culture were mixed in each well and then kept at room temperature for 30 min. The detection solution was prepared by mixing equal volume of 2 mg ml−1 nitro blue tetrazolium and 0.1 mg ml−1 phenazine ethosulfate and 20 µl of the solution was added to each well. After the incubation at room temperature in the dark for 90 min, absorbance at 660 nm was analyzed and IC50s were calculated from dose response curve.