Reaction hijacking of tyrosine tRNA synthetase as a new whole-of-life-cycle antimalarial strategy.

Aminoacyl transfer RNA (tRNA) synthetases (aaRSs) are attractive drug targets, and we present class I and II aaRSs as previously unrecognized targets for adenosine 5'-monophosphate-mimicking nucleoside sulfamates. The target enzyme catalyzes the formation of an inhibitory amino acid-sulfamate conjugate through a reaction-hijacking mechanism. We identified adenosine 5'-sulfamate as a broad-specificity compound that hijacks a range of aaRSs and ML901 as a specific reagent a specific reagent that hijacks a single aaRS in the malaria parasite Plasmodium falciparum, namely tyrosine RS (PfYRS). ML901 exerts whole-life-cycle-killing activity with low nanomolar potency and single-dose efficacy in a mouse model of malaria. X-ray crystallographic studies of plasmodium and human YRSs reveal differential flexibility of a loop over the catalytic site that underpins differential susceptibility to reaction hijacking by ML901.

Diseases caused by infectious organisms pose an enormous threat to global health, food security and sustainable development. Malaria is one such debilitating disease, caused by protist parasites of the genus Plasmodium. Every year at least 200 million new infections of P. falciparum malaria are established, causing more than 600,000 deaths (1). Current antimalarial treatments are rapidly losing efficacy; and standard-of-care artemisinin combination therapies fail to cure infections in ~50% of patients in some regions of Asia (2). Clinically validated resistance to artemisinins has now been detected in Africa (3), where most malaria deaths occur. New treatments with novel modes of action are urgently needed to overcome existing resistance, expand possible treatment options and enable more effective combination therapies.

Adenosine 5'-sulfamate exhibits broad specificity reaction hijacking, revealing potential antimalarial drug targets

Nucleoside sulfamates, such as the investigational drug Pevonedostat (4), inhibit ubiquitin-like protein (UBL) activating enzymes (E1s) by forming covalent conjugate inhibitors with the enzyme-bound UBL. The E1s catalyze nucleophilic attack of the sulfamate nitrogen on the thio-ester bond between the UBL and the E1 (Fig. 1A, fig. S1). Until now, attack on the thio-ester bonds of UBLs was the only known example of this type of inhibitor mechanism. However, naturally occurring nucleoside sulfamates and derivatives, such as nucleocidin (5), 2-Cl-adenosine sulfamate (6, 7) and adenosine 5′-sulfamate (AMS) (8), exhibit inhibitory activity against bacteria (6–8), which lack E1 enzymes. The compounds are broadly toxic and have been reported to inhibit protein synthesis (8, 9), but until now the mechanisms underlying these activities were unknown.

Figure 1

AMS-treated infected RBCs reveals aaRSs as potential targets.

(A) E1 enzymes can catalyze attack of the sulfamate nitrogen on the carbonyl carbon of the thioester bond between the UBL and the E1 to form a UBL conjugate. aaRSs could catalyze nucleoside sulfamate attack on activated amino acids to form an amino acid adduct. (B) Structure of 5′-adenylate sulfamate. (C) Trophozoite stage parasites were incubated with DMSO (Mock), different concentrations of AMS, or, borrelidin (BOR; a threonyl-tRNA synthetase inhibitor). Western blots of lysates were probed for phosphorylated-eIF2α with PfBiP as a loading control. The blot is typical of data from three independent experiments. (D) P. falciparum-infected RBCs were treated with 10 μM AMS for 3 h. Extracts were subjected to LCMS analysis identifying the Tyr-AMS conjugate. The profile is typical of data from three independent experiments.

We explored the activity of adenosine 5′-sulfamate (AMS, Fig. 1B), a close mimic of AMP, as a potential starting point for identifying antimalarial compounds. We found that AMS is highly cytotoxic (IC50_72h = 1.8 nM) to P. falciparum cultures, with an efficacy similar to that of current front-line drug, dihydroxyartemisinin (DHA), but is also cytotoxic to mammalian cell lines, such as HCT116 (IC50_72h = 26 nM) (table S1). We found that treatment of P. falciparum cultures with AMS triggers eIF2α phosphorylation (Fig. 1C), a hallmark of stress caused by either accumulation of unfolded proteins or uncharged tRNAs (10). Like E1 enzymes, aminoacyl tRNA synthetases (aaRSs) are adenylate-forming enzymes (AFEs). They catalyze the transformation of amino acids into AMP conjugates, and then into aminoacyl-tRNAs, to supply protein synthesis. Given the reported effects on protein translation (8, 9), we considered the possibility that aaRSs might be able to catalyze nucleophilic attack of AMS on their cognate aminoacyl tRNAs (Fig. 1A).

The proposed mechanism would be expected to generate AMS-amino acid conjugates (Fig. 1A), so we used targeted mass spectrometry to search for the predicted conjugates in P. falciparum-infected red blood cells (RBCs) and cultured human cells (HeLa) that had been treated with 10 μM AMS for 2-3 h (see Supplementary Material for full methods). Following Folch extraction of lysates, the aqueous phase was subjected to liquid chromatography-coupled mass spectrometry (LCMS) and the anticipated masses for the 20 possible amino acid conjugates were interrogated. In P. falciparum, the extracts yielded a strong signal for AMS-Tyr (Fig. 1D), with matching precursor ion m/z (< 3 ppm) and anticipated fragmentation spectrum (fig. S2A). MS peaks were also detected for the adducts of Asn, Asp, Ser, Thr, Gly, Ala, Lys and Pro (fig. S2B). In the mammalian cell line, AMS conjugates were identified for Asn, Pro, Ala, Thr, Asp and Tyr (fig. S3). No peaks were detected in control samples. These data are consistent with aaRSs catalyzing nucleoside sulfamate attack on the activated oxy-ester bonds of their cognate aminoacyl tRNAs (Fig. 1A), Thus, both class I and class II aaRSs are potentially susceptible to inhibition via the reaction hijacking mechanism.

Identifying a nucleoside sulfamate with potent and specific antimalarial activity

In an effort to identify aaRS-targeting nucleoside sulfamates with narrower specificity, we screened 2314 sulfamates from the Takeda compound library (Cambridge, MA, USA) for inhibition of the growth of P. falciparum. The library included compounds that were synthesized as potential inhibitors of Atg7, an E1 that activates UBLs, including the Atg8s (11). We identified several pyrazolopyrimidine sulfamates with a 7-position substituent (exemplar ML901; Fig. 2A) that possess potent activity against P. falciparum. The ML901 50% inhibitory concentration (IC50_72h = 2.0 ± 0.1 nM) is similar to that for DHA (table S1).

Figure 2

ML901 exhibits potent activity against P. falciparum in vivo.

(A) Structure of pyrazolopyrimidine ribose sulfamate, ML901. (B) Pharmacokinetics profile (in blood) over the first day for SCID mice engrafted with human RBCs infected with P. falciparum following treatment with ML901 at 50 mg/kg i.p. (C) Therapeutic efficacy of ML901 in the SCID mouse P. falciparum model, dosed with ML901 at 50 mg/kg i.p. in comparison with gold standard antimalarial, chloroquine, dosed at 50 mg/kg p.o..

ML901 was tested for cytotoxicity against different mammalian cell lines (table S1) and showed 800- to 5,000-fold selectivity towards P. falciparum (>1,000-fold higher selectivity than AMS). ML901 retained activity against all strains of P. falciparum tested, regardless of their resistance profile and geographical origin (table S2). It potently inhibited transmissible male gametes (table S2); and prevented development of P. falciparum in primary human hepatocytes (table S2). We confirmed that ML901 exerts activity against human Atg7 (IC50 = 33 nM); but has much weaker activity against other E1 enzymes (table S3), as previously reported for nucleoside sulfamates with a substitution at the 7-position (11); and consistent with the low mammalian cell cytotoxicity. By contrast, AMS is a potent inhibitor of each of the E1s tested (table S3). The rat pharmacokinetic profile of ML901 (fig. S4; table S3) is encouraging, with low blood clearance and a long terminal half-life in blood (T1/2∞ = 41 h) following intravenous or oral dosing.

We determined the in vivo antimalarial efficacy of ML901 in severe combined immune deficient (SCID) mice, engrafted with P. falciparum infected human RBCs (12, 13), which is the gold standard for testing in vivo efficacy of malaria drug candidates. A single dose (50 mg/kg i.p.) results in excellent exposure (area under the curve = 580 μM.h; Fig. 2B) and achieves reduction of parasitemia to baseline (Fig. 2C), with no evidence of toxicity. The clearance rate is similar to that of chloroquine (CQ; 50 mg/kg p.o.).

ML901 selectively targets plasmodium tyrosine tRNA synthetase

To identify the target of ML901 in P. falciparum, extracts of infected RBCs that had been treated with 3 μM ML901 (3 h) were subjected to LCMS to search for amino acid-ML901 conjugates. An LCMS peak corresponding to protonated ML901-Tyr (m/z = 576.1324) was detected. Synthetic ML901-Tyr was generated and spiked into the untreated P. falciparum lysate to confirm the peak assignment (fig. S5). None of the other 19 possible amino acid conjugates were detected.

To determine whether P. falciparum tyrosine tRNA synthetase (PfYRS) is the critical target in P. falciparum, we subjected parasite cultures (3D7 line) to increasing concentrations of the compound over a period of 4 months; retrieving parasites with 10-fold reduced sensitivity (ML901 IC50_72h = 28 nM) (fig. S6A). Whole genome sequencing of one parental and three resistant clones revealed nine newly acquired mutations in the resistant clones, three of which were in PF3D7_0807900 (position 403,556), corresponding to a Ser234Cys mutation in cytoplasmic PfYRS (table S4). Insertion of an additional Asn in an Asn repeat in kinesin-5 is considered unlikely to be functionally significant (see table S4). No mutations were observed in any other AFE, including P. falciparum Atg7. Transfectants were generated (Dd2 parent) harboring the PfYRSS234C mutation (fig. S6B-D), which recapitulated the resistance phenotype (i.e., 10-fold decreased sensitivity) (Fig. 3A; table S5). Aptamer-induced downmodulation of PfYRS decreased the growth rate compared with the control (fig. S6E) and increased sensitivity to ML901 (Fig. 3B; table S5); but not to DHA or Thr-RS inhibitor, borrelidin (fig S6F; table S5). The remarkable potency of ML901 against the knockdown parasites (IC50 = 0.4 nM) both validates PfYRS as the target and points to an extremely potent inhibitory interaction.

Figure 3

ML901 targets PfYRS and inhibits protein translation.

(A,B) Sensitivity to ML901 exposure (72-h) for a cloned wildtype line (Dd2) and 3 CRISPR-edited clones harboring PfYRSS234C (A) or an aptamer-regulatable PfYRS line upon addition of aTc, with data normalized to a no drug control (B). See table S5 for data values. (C) RBCs infected with schizont stage (43-46 h p.i.) P. falciparum (Cam3.II-rev) were exposed to ML901 for 3 h. Protein translation was assessed in the second two hours of the incubation, via the incorporation of OPP. Aliquots of inhibitor-exposed cultures were washed and returned to cultures, and viability was estimated at the trophozoite stage of the next cycle. IC50 (Translation) = 65 nM, IC50 (Viability) = 56 nM. Data are representative of three independent experiments. Error bars correspond to the range of technical duplicates. (D) Schizont stage Cam3.II_rev parasites were incubated with DMSO (Mock), 1 μM DHA, 200 nM borrelidin (BOR) or 200 nM ML901 for 3 h and Western blots of lysates were probed for phosphorylated-eIF2α with PfBiP as a loading control. The blot is typical of data from three independent experiments.

ML901 inhibits protein translation in P. falciparum schizonts (as monitored by O-propargyl-puromycin incorporation (14); Fig. 3C), consistent with PfYRS being the target. The IC50_3h value (50 nM) correlates well with that for parasite killing potency (Fig. 3C). Another protein translation inhibitor, cycloheximide, has a similar profile (fig. S7A), while the folate pathway inhibitor, WR99210, kills parasites with no immediate effect on protein translation (fig. S7B). ML901 triggers eIF2α phosphorylation in wildtype P. falciparum (Cam3.II_rev; Fig. 3D), consistent with the presence of uncharged tRNA (10, 15). In eIK1 (GCN2 equivalent (16)) knockout parasites, the amino acid starvation pathway is disrupted and ML901 treatment does not result in eIF2α phosphorylation (fig. S7C), consistent with an aaRS target (fig. S7D).

YRSs from wildtype (PfYRS), mutant (PfYRSS234C) and human (mature HsYRS (17)) were produced in Escherichia coli. Biophysical characterization revealed well-folded dimers (fig. S8,9; table S6). PftRNATyr and HstRNATyr (18) were generated by in vitro transcription (fig. S10). When ML901 was incubated with PfYRS in the presence of all other substrates (i.e., Tyr, ATP and tRNATyr), the apparent protein melting point (Tm), measured by differential scanning fluorimetry (DSF), increased by a remarkable 15°C (Fig. 4A,B; fig. S11A). The increase in thermal stability is even greater than that induced by the tightly bound adenylate intermediate, AMP-Tyr (table S7). Whereas formation of the AMP adenylate requires only Tyr and ATP, the thermal stabilization induced by ML901 required all three substrates (Tyr, ATP and tRNATyr). This result is consistent with a hijacking mechanism that requires charged tRNATyr (Tyr-tRNATyr; see Fig. 1A). Importantly, recombinant HsYRS was not stabilized in the presence of ML901 plus substrates, suggesting the inhibitory species is not produced by, or does not bind to, the human enzyme (Fig. 4B, red curves). The mutant PfYRSS234C was less well stabilized (fig. S11A, table S7), suggesting weaker binding of the inhibitory species, consistent with the mutant parasite’s resistance to ML901.

Figure 4

ML901 inhibits PfYRS by a reaction-hijacking mechanism.

(A) The apparent melting temperature (Tm) of PfYRS after incubation at 37°C for 3 h with the indicated reactants: ML901 (50 μM), ATP (50 μM), tyrosine (100 μM), PftRNATyr (4 μM). Data represent the average of three independent assays and error bars correspond to SD. (B) First derivatives of melting curves for PfYRS and HsYRS with or without pre-incubation with ML901 (50 μM), ATP (50 μM), tyrosine (100 μM) and PftRNATyr (4 μM). Data is representative of three independent assays. (C) Effects of increasing concentrations of ML901 on tyrosine acylation of the cognate tRNATyr by PfYRS and HsYRS with YRS (0.25 μM), ATP (10 μM), tyrosine (100 – 200 μM), cognate tRNATyr (24 μM) and pyrophosphatase (1 unit/mL), at 37°C for 1 h. IC50 (PfYRS) = 53 μM; IC50 (HsYRS) > 500 μM. Data represent the average of 8 independent assays and error bars correspond to SEM. (D) Structure of ML901-Tyr. (E) PfYRS was incubated with ML901 (50 μM), ATP (10 μM), tyrosine (20 μM) and PftRNATyr (8 μM). Following urea denaturation and TFA precipitation, the supernatant was subjected to LCMS analysis, revealing the expected protonated ML901-Tyr ion. The profile is typical of data from three independent experiments. (F) Schematic of reaction-hijacking mechanism.

ML901 exerts its activity by hijacking the active site-bound reaction product

We examined the abilities of the recombinant YRSs to consume ATP, i.e., to form and release AMP-Tyr in the initial reaction phase. HsYRS and PfYRSS234C show higher activity (in the absence of tRNA) than PfYRS (fig. S11B). This difference suggests that AMP-Tyr is bound less tightly to the Hs and mutant PfYRS active sites. Upon addition of the cognate tRNATyr, ATP consumption is enhanced, consistent with productive aminoacylation. Acylation of the cognate tRNATyr to radiolabeled tyrosine (19), occurs at a similar level in all three enzymes (fig. S11C). ML901 inhibits ATP consumption by PfYRS when added in the presence of PftRNATyr but not in its absence (fig. S11D). Similarly, ML901 inhibits tRNATyr acylation to tyrosine in vitro by PfYRS, but not by HsYRS (Fig. 4C). Synthetically generated ML901-Tyr conjugate is able to inhibit the activity of both PfYRS and HsYRS (fig. S11E,F), suggesting that, although HsYRS is unable to generate the ML901-Tyr inhibitor, it can bind the preformed conjugate.

To confirm that recombinant PfYRS can generate the ML901-Tyr conjugate, we incubated PfYRS with substrates and ML901, then precipitated the tRNA and protein. The supernatant was subjected to LCMS analysis and we detected a peak at m/z 576.1331 Da (Fig. 4D,E), consistent with the expected protonated ML901-Tyr ion, and confirmed using the synthetic ML901-Tyr standard (fig. S12A-C). Under the conditions used, PfYRS generates more ML901-Tyr than PfYRSS234C dimer (fig. S12E); and no ML901-Tyr was detected when HsYRS was incubated with ML901 and substrates. Thus, it appears that the ability of PfYRS to catalyze formation of the ML901-Tyr conjugate is the primary factor controlling selectivity versus HsYRS and potency versus PfYRSS234C.

A structured loop over the PfYRS active site facilitates reaction hijacking

To obtain crystals of ML901-Tyr-bound PfYRS, we incubated His-tagged PfYRS with tyrosine, ATP and ML901 in the presence of tRNA and then purified the complex using nickel affinity and gel filtration chromatography. The crystal structure (refined at 2.15-Å resolution) revealed a homodimer organization with clear density for the ML901-Tyr ligand (Fig. 5A, fig. S13A).

Figure 5

Structural analysis of YRSs reveals the determinants of potency and specificity.

(A) The structure of the dimeric PfYRS/ML901-Tyr complex showing chain A (green), chain B (blue), and bound ML901-Tyr (red stick representation). (B) Inhibitor/active site interactions for the B chain. (C) (i) The PfYRS chain B active site highlighting the “HIGH” (70HIAQ73; light purple) and “KMSKS” (247KMSKS251; light brown) motifs with bound ML901-Tyr (colored by atom type). M248 and H70 are positioned to interact. (ii) Active site of HsYRS with bound ML901-Tyr highlighting the “HIGH” motif (49HVAY52; light pink). (iii) Active site of PfYRSS234C with bound ML901-Tyr highlighting the “HIGH” (70HIAQ73; light green). Unmodelled loops are shown in Cii and Ciii as dashed lines. (D) Overlay of PfYRS (B chain), HsYRS, and PfYRSS234C showing the different configurations adopted by His70/His49. PfYRS His70, purple; HsYRS His49, pink; PfYRSS234C His70, green.

PfYRS is a Class I aaRS, characterized by a catalytic domain that adopts a Rossmann fold (residues 18–260) linked to a C-terminal domain (residues 261–370) that is involved in recognition of the anticodon stem of tRNATyr. PfYRS contains the two motifs characteristic of the catalytic domain of Class I (sub-class c) tRNA synthetases: “HIGH” and “KMSKS” (70HIAQ73 and 247KMSKS251 in PfYRS). The overall structure of the ML901-Tyr/PfYRS complex is very similar to our structure of PfYRS with the native ligand, AMP-Tyr (fig. S14, table S8) and the previously published structure (19).

ML901-Tyr makes many interactions with active site residues, involving the pyrazolopyrimidine amine, ribose hydroxyls, sulfamate, and tyrosine (Fig. 5B; fig. S13B), which underpin the tight binding affinity; and potency. ML901-Tyr is present in the active sites of both monomers in the dimer (fig. S13C,D,J,K), but differences are observed with respect to the KMSKS loop, which forms a flap over the adenylate/ML901 binding site (Fig. 5B,Ci, fig. S13C-F). In the B chain, His70 (of 70HIAQ73) makes close contact with Met248 in the KMSKS loop (Fig. 5Ci) and is well defined in the electron density (fig. S13F). In the A chain, part of the KMSKS loop is not well defined (fig. S13E), suggesting that the A chain loop is more mobile, and His70 adopts a side chain rotamer that would clash with Met248 if the loop was structured as in chain B (fig. S13F-I). In combination, these observations indicate different conformations of the KMSKS loop in individual monomers within the dimer. By contrast, the KMSKS loops of both monomers of AMP-Tyr-bound PfYRS (fig. S14) are well defined, with the electron density clearly showing contacts between His70 and Met248 in both subunits.

In the published structure of tyrosine-bound HsYRS (PDB: 4QBT (20)) the equivalent “KMSSS” loop is not modelled, suggesting that it is mobile. We were not able to generate a structure of HsYRS in complex with enzyme-generated ML901-Tyr as HsYRS does not catalyze formation of the conjugate. However, we were able to form crystals of HsYRS in the presence of synthetic ML901-Tyr (fig. S15), consistent with our observation that the preformed conjugate can inhibit HsYRS activity (fig. S11F). Although we observed clear density for ML901-Tyr (fig. S15F), the 222KMSSS226 loops were not defined in the electron density (Fig. 5Cii; fig. S15D,E). Moreover, His49 (the equivalent of PfYRS His70) adopts a position (Fig. 5D, magenta) similar to that in chain A of ML901-Tyr-bound PfYRS (compare figures S13G,H and S15E), further suggesting that this configuration is associated with increased loop mobility. A comparison of the interaction networks (fig. S13B-D, S15B,C) reveals notably fewer interactions with the ML901 moiety in HsYRS compared with PfYRS; and specific interactions were poorly conserved between the two enzymes.

We also solved the structure of PfYRSS234C in complex with synthetic ML901-Tyr (fig. S16). Similar to HsYRS, the KMSKS loops of both monomers were not defined in the electron density (fig. S16E,F), and His70 adopts a rotamer that is not consistent with a structured KMSKS loop (Fig. 5Ciii,D, green; compare figures S13G,H and S16G,H). Potency and selectivity of ML901 for PfYRS thus appears to be associated with a stabilized loop over the active site. That is, the decreased susceptibility of HsYRS and PfYRSS234C to reaction hijacking by ML901 is associated with mobility of the KMSSS/KMSKS loop, which is, in turn, associated with rotation of the His49/70 side chain. These conformational changes may promote dissociation of the charged tRNA, thereby preventing the hijacking reaction.

The pyrazolopyrimidine sulfamate chemotype is an attractive starting point for a malaria drug discovery program, based on our observation that the specific inhibition of PfYRS by ML901 is lethal to disease-causing and transmissible stages of P. falciparum, and that ML901 exhibits a long in vivo half-life, underpinning its single-dose efficacy in a murine model of human malaria. Further exploration of substitutions at the 7-position of the pyrazolopyrimidine sulfamates class is expected to identify compounds with reduced activity against human Atg7, and thus even higher specificity for plasmodium. We note that the HIAQ and KMSKS motifs are conserved across apicomplexan and kinetoplastid parasites but not in metazoan organisms (fig. S17). This suggests that ML901-like compounds could exhibit cross-pathogen inhibitory activity. Use of the sulfamates in a drug combination could prevent evolution of resistant mutants.

Our finding that nucleoside sulfamates can hijack Class I and Class II tRNA aaRSs, as well as E1s, opens up the possibility of designing bespoke small molecular weight, membrane permeable AFE inhibitors with adjustable specificity. In addition to charging tRNA and activating ubiquitin, AFEs are involved in activating fatty acids for degradation, biosynthesis of natural products, and other diverse pathways (21). Thus, nucleoside sulfamates may find applications in a broad range of infectious, metabolic and neurodegenerative diseases.