Optimization of Methionyl tRNA-Synthetase Inhibitors for Treatment of Cryptosporidium Infection

Although the world has seen substantial progress in the reduction of child mortality in recent decades, much work remains to achieve the United Nations’ Sustainable Development Goals (targeting 2030) for child survival (1). Diarrhea is one of the leading causes of child mortality, responsible for 8% of deaths in children aged 1 to 59 months (2). The Global Enteric Multicenter Study (GEMS) identified Cryptosporidium to be the second leading cause of moderate to severe diarrhea in young children at sites in Africa and Asia (3). The importance of cryptosporidiosis in community diarrhea in developing countries was confirmed in the MAL-ED study (4). Beyond the mortality risk, children who survive cryptosporidiosis suffer from growth and developmental stunting, which contribute to all-cause mortality and disability (5, 6). The recent appreciation of the impact of cryptosporidiosis has drawn attention to the inadequacies in the means to control this infectious disease. No vaccines are in clinical use, and the sole drug for treating cryptosporidiosis (nitazoxanide) has poor efficacy in malnourished children and in patients with human immunodeficiency virus. Vaccine development is likely to be slow due to difficulties raised by antigenic differences within Cryptosporidium, leading to poor cross-protection between species and strains (7). New anticryptosporidial drugs are likely to be the most rapidly developed technology to address the burden of Cryptosporidium infection.

Recent studies suggest that Cryptosporidium is more closely related to gregarine protozoa than to coccidians (8). The genus Cryptosporidium has 27 species that have been identified worldwide that infect four classes of vertebrates (9). The species primarily responsible for human cryptosporidiosis are Cryptosporidium hominis and Cryptosporidium parvum. Water- and foodborne transmission are the major modes of infection, although person-to-person contact is also described (9). The small intestine is the primary site of Cryptosporidium infection in humans, although extraintestinal sites such as the biliary tract, lungs, and pancreas can be involved in immunocompromised and immunocompetent individuals (9–11). The extraintestinal locations may have important implications for developing therapeutics that act at all sites of infection. Cryptosporidium mainly resides in an unusual niche in the intestinal epithelium known as the parasitophorous vacuole, which is insulated from both the intestinal lumen, as well as the host cytoplasm. Therefore, it is not entirely known whether anti-Cryptosporidium drugs should be optimized for luminal or plasma exposure, although a recent study emphasizes the importance of gastrointestinal (luminal) exposure in a murine model (12).

Protein synthesis is a classic antimicrobial drug target, dating back to many of the first antibiotics—such as chloramphenicol, tetracycline, and erythromycin—that inhibit bacterial protein synthesis. Recently, aminoacyl-tRNA synthetase (aaRS) inhibitors have emerged as promising therapeutic candidates for targeting protein synthesis. Using ATP hydrolysis, aaRSs catalyze the formation of tRNAs charged with their cognate amino acids, which serve as the substrates for the formation of new peptides. Mupirocin, a small molecule inhibitor of isoleucyl-tRNA synthetase (13), has been in clinical use for more than 2 decades as a topical treatment for Staphylococcus infections. Tavaborole is a leucyl-tRNA synthetase inhibitor that was approved by the U.S. Food and Drug Administration in 2014 for topical treatment of onychomycosis (14, 15). Halofuginone, a prolyl-tRNA synthetase inhibitor (16), is approved for veterinary use against Cryptosporidium in Europe, although a narrow therapeutic index makes it unsuitable for human use. Three different aaRS inhibitors for systemic use are in clinical trials, demonstrating the potential for safe use beyond topical applications. These include the methionyl-tRNA synthetase (MetRS) inhibitor CRS3213 for Clostridium difficile infections (ClinicalTrials.gov, NCT01551004), the leucyl-tRNA synthetase inhibitor GSK2251052 for Gram-negative bacterial infections (17, 18), and the leucyl-tRNA synthetase inhibitor GSK3936656 for multidrug-resistant tuberculosis (19, 20). Inhibitors of other aaRSs from protozoan parasites, including Cryptosporidium, Plasmodium, Trypanosoma, and Toxoplasma, are also in development (21, 22).

MetRS enzymes fall into two categories: MetRS1 and MetRS2 (23). C. parvum and C. hominis contain a single MetRS gene that aligns with the MetRS1 category, meaning it has close homology to the MetRS of S. aureus, Trypanosoma spp., and the human mitochondrial MetRS. Our group has been developing inhibitors to type 1 MetRS that are shown to have potent activity against trypanosomes and Gram-positive bacteria, including activity in animal models (24–27). Supporting this work are numerous crystal structures of the inhibitors bound to the Trypanosoma brucei MetRS enzyme (28–30), which is 76% identical (19 of 25 residues) to the C. parvum/C. hominis MetRS within the inhibitor binding pocket. Here, the C. parvum MetRS was characterized and MetRS inhibitors were shown to be highly potent with K_i values as low as 0.9 pM. Compound 2093 had the most potent in vitro activity against C. parvum and reduced Cryptosporidium infection to low levels in two murine models without showing signs of toxicity. The physicochemical and pharmacological features associated with in vivo activity are discussed. The research illustrates the potential of target-based drug discovery to develop a novel therapeutic against a formidable eukaryotic pathogen.

The Cryptosporidium enzyme, methionyl-tRNA synthetase, was targeted for the development of novel drugs to treat cryptosporidiosis. Analysis of the deposited amino acid sequence of the CpMetRS revealed that it clusters with type 1 MetRS enzymes (23) that include trypanosomes, Giardia, and Gram-positive bacteria. In previous work, crystal structures of the T. brucei MetRS revealed the ATP and methionine binding pockets, as well as the binding mode for numerous inhibitors (28, 29). By mapping the C. parvum enzyme to the T. brucei MetRS structure, the analogous binding pocket was identified, demonstrating that 19 of 25 amino acids (76%) forming the surface of the pocket are identical. This indicated that inhibitors of the T. brucei MetRS would be likely to bind the CpMetRS. The binding pocket residues of the C. hominis MetRS are identical to those of C. parvum. The sequence analysis also showed that 18 of 24 residues (75%) were identical in the corresponding human mitochondrial MetRS. The significance of this similarity is discussed below. In contrast, the human cytoplasmic MetRS had a lower degree of identity, i.e., 14 of 24 (58%), which is consistent with the knowledge that the human cytoplasmic MetRS belongs to the MetRS2 category which is not inhibited by the compounds under investigation.

The CpMetRS gene was amplified from genomic DNA, cloned into an expression vector, and overexpressed in E. coli. The purified enzyme was catalytically active in an aminoacylation assay in which [³H]methionine is incorporated into tRNA (Fig. S2). This method uses 100 nM enzyme to provide an acceptable signal to background ratio which constrains the ability to accurately measure the K_i for highly potent inhibitors (since the 50% inhibitory concentration [IC₅₀] can theoretically be no less than half the enzyme concentration). In order to measure the K_i for the MetRS inhibitors, the ATP:PP_i exchange assay was adopted (31). In this method, incorporation of [³²P]PP_i into ATP occurs by the reverse enzyme reaction that can be measured when the receiving substrate, tRNA, is omitted. The K_m for the enzyme substrates, methionine and ATP, are within a factor of 2 to 3 of the values measured in this study and reported elsewhere (31) for the S. aureus MetRS. In contrast, the K_m values of methionine and ATP for the human mitochondrial MetRS are lower by a factor of 5 to 10.

Since the MetRS inhibitors are competitive with methionine (31), it is possible to accurately measure the IC₅₀s of inhibitors in the reaction by raising the concentration of methionine above its K_m (34). The shift in IC₅₀ is proportional to the [substrate]/K_m, as described by the Cheng-Prusoff equation for competitive inhibitors (35). Of note, the method of raising the methionine substrate concentration cannot be used in the aminoacylation assay mentioned earlier because the unlabeled methionine competes with the [³H]methionine that is necessary for the readout. The K_i values were determined for selected MetRS inhibitors revealing extraordinary potency of many of these compounds, ranging from 0.0009 to >1.33 nM (Table 4). This is similar to the K_i for antibiotic MetRS inhibitor, REP8839 against the S. aureus MetRS, reported at 0.01 nM (31). The relationship between the K_i and C. parvum EC₅₀ showed a strong correlation (R² = 0.91; P < 0.0001), which is consistent with the observation that compound 2093 was the most potent compound against the MetRS enzyme and against C. parvum infection of cell cultures. The correlation data support the conclusion that the inhibitors act “on target” to mediate their effects on C. parvum cells.

More than 500 MetRS inhibitors have been developed in our program to optimize their antitrypanosomal and antibacterial activities (24–26, 30). A set of structurally diverse compounds from this library (Tables 3 and 4) was screened against C. parvum, revealing a mix of positive and negative results. The aminoquinolone-containing compounds (exemplified by compound 1312 [24]) had poor activity. The aminoquinolone compounds are in clinical development as antibiotics for Clostridium difficile and S. aureus infections (36, 37). Similarly, compounds with the urea moiety (e.g., compound 1433 [25]) also had poor activity. In contrast, compounds with the fluoro-imidazopyridine (e.g., compounds 1614 and 1717 [26]) demonstrated more potent activity, in the 5 to 10 μM range. Parallel work had indicated that changes to the linker region of the molecule were well tolerated, leading to explorations of various changes, including ring systems at this region of the molecule (30). Compounds with the 1,3-dihydro-2-oxo-imidazole as the linker ring were particularly active (Table 4). Among the compounds containing the 1,3-dihydro-2-oxo-imidazole, various substitutions were explored at the R groups indicated in Table 4. R₁ as ethyl (e.g., compound 2062) was more active than R as H (compound 1962) or as propyl (compound 2069). R1 as CH₂CF₂H appears to be slightly less active than the ethyl version. R₂ as F is slightly less active than R₂ as Cl, and Br is essentially the same as Cl. Next, substitutions on the benzyl group (on the left side of the structure; Table 4) were explored. It had been previously shown that 3,5 substitutions (such as 3,5-dichlorobenzyl) were particularly potent on the T. brucei MetRS (24). In the orientation created by the ring-linker structures, the 2,4 substitutions (as shown in Table 4) have the greatest potency. Changing R₄ from Cl to OCH₃, produced the most potent inhibitors in the series (e.g., compounds 2093 and 2114). Compounds with a trisubstituted benzyl group (e.g., 2,4,5 substitutions) had much diminished activity against C. parvum oocysts (data not shown). In summary, compound 2093 was the most potent compound (EC₅₀ = 0.007 μM) and compares very favorably to the published data for the clinical drug, nitazoxanide (EC₅₀ = 3.7 μM) against C. parvum (38).

The activity of the MetRS inhibitors against C. parvum infection was tested in two different murine models. The NOD SCID gamma mouse model was performed with a PCR-based readout comparing pretreatment (day 5) to posttreatment (day 11) fecal parasite levels. Fecal oocysts were significantly reduced for compound 2093 (tested twice), as well as for compounds 2069 and 2259, but not for compound 2067. The activities were then retested in the Cryptosporidium infection model using adult IFN-γ KO mice. In this model, the mice were monitored for 20 days postinfection. Quantitation of stool parasite loads again showed that compound 2093 was highly efficacious (>4-log₁₀ drop in luminescence) at an oral dose of 50 mg/kg BID (Fig. 3A). The greater magnitude of parasite reduction observed in the IFN-γ KO mice may be due to differences in the models and readout methods. In addition, the challenge dose was lower in the IFN-γ KO mice (10³) compared to the NSG model (10⁵), which may account for the differences if there is an inoculum effect as seen with some bacterial infections (39). A persistence of a low luminescence signal above background levels (log₁₀ RLU of 2.5) was detected during the remainder of the monitoring period in the IFN-γ model, the significance of which is unclear. Importantly, the parasite signal did not rebound to the high levels observed with the controls. Also, the mice maintained their body weight and survived to the end of the 20-day observation period, unlike the vehicle-treated mice that needed to be euthanized on day 17 due to the loss of >20% body weight (Fig. 3A). Several follow-up experiments confirmed the in vivo activity of compound 2093 at lower doses (e.g., 50 mg/kg/day divided into one or two doses), but the activity was substantially diminished at the dose of 20 mg/kg once per day. The other compounds with excellent in vivo activity were compounds 2114 and 2259. The mice appeared to tolerate the treatments without any observed side effects and did not experience the weight loss that was observed in the vehicle-treated mice. The one exception was for compound 2258, where the mice lost weight and were euthanized on day 13 (before the control mice needed to be euthanized). For the other compounds, it was encouraging to observe that potent anticryptosporidial activity was associated with good clinical outcomes in the mice.

The compounds with the greatest in vivo efficacy (compounds 2093, 2114, and 2259) were among the most potent in vitro compounds. To further understand the reasons for the more potent in vivo activity, various physicochemical properties were assessed. Some general characteristics of these potent in vivo compounds are molecular weights in the 400 to 450 range, log P values in the 3.5 to 3.75 range, and a relatively high solubility (>25 μM) at pH levels reflective of the gastrointestinal tract and plasma. The MetRS inhibitors are all highly protein-bound in the range of 95 to 99.9% (except for compound 2240 with 87.7% protein binding). Apparently, high protein binding is not detrimental to activity since compound 2093 is 99.9% bound to mouse plasma proteins. Figure 4 allows for visualization of the chemical properties as they relate to in vivo efficacy. It shows that the most active compounds in vivo have relatively high solubility/EC₅₀, apparently reflecting the importance of in vitro potency (EC₅₀) and, at least, moderate solubility. Interestingly, the most effective compounds had intermediate levels of predicated permeability perhaps, suggesting that overly permeable compounds may be completely absorbed into the blood and thus unavailable in the gut for local antiparasitic activity and that compounds with low permeability may not cross membranes sufficiently to exert effects on the parasites either.

The liver microsome metabolic half-lives for the MetRS inhibitors were relatively low (<15 min for most compounds tested; Table 6). In fact, the lead compound 2093 has half-lives of 3.2 and 8.3 min in murine and human microsomes, respectively. This short in vitro half-life does not directly translate to low plasma exposure in vivo, probably because of the protective effects of the high plasma protein binding.

The pharmacokinetic (PK) properties in mice of several of the MetRS inhibitors were also assessed. The two most potent compounds (2093 and 2114) had similar PK profiles, with a C_max in the 6 to 8 μM range and an AUC of ∼2,000 min ⋅ μmol/liter. The fecal levels of compounds 2093 and 2114 were 31 and 11 μM, respectively. The fecal levels suggest that sufficient amounts of intact compound (>100 × EC₅₀) are available in the fecal stream to exert anticryptosporidiosis effects. Previous studies with “bumped kinase inhibitors” indicate that intestinal levels of compound better correlate with antcryptosporidiosis activity than do plasma levels (12). Both fecal levels and plasma levels were similar enough among the tested CpMetRS inhibitors; thus, it is not possible to make firm conclusions about the most favorable properties. One exception is that compound 2240 had very high average fecal levels (846 μM) and yet had no in vivo anti-Cryptosporidium activity. We speculate that this compound may have passed through the gastrointestinal tract in an insoluble form that was not available for local anticryptosporidiosis activity. If this is true, then merely delivering insoluble compound to the gut is not sufficient. Since the pathogen lives in an intracellular niche, it is necessary for the compound to be in solution and sufficiently permeable to reach that niche either directly from the intestinal lumen or via the bloodstream.

The safety of the compounds is of great importance given that the target population for anticryptosporidiosis treatment will include very young children and other vulnerable groups. A potential concern for MetRS inhibitors is cross-activity on the human mitochondrial MetRS enzyme that could lead to mitochondrial dysfunction. The human mitochondrial MetRS and the CpMetRS are identical at 18 of 25 residues in the compound binding pocket (see Table 1), indicating moderate similarity. An assay was performed to quantify cytochrome oxidase 1 (COX-1) enzyme levels in human liver cells (HepG2) after a 6-day incubation with MetRS inhibitors. This enzyme is encoded and expressed in the mitochondrion, whereas the control protein (SDH-A) is encoded in the nucleus and expressed in the cytoplasm. The single ring-linker MetRS inhibitors (Table 4), in fact, demonstrated substantial inhibition of COX-1 expression levels with EC₅₀ values as low as 0.039 μM in the case of compound 2093. This is only slightly above the EC₅₀ value against C. parvum (0.007 to 0.036 μM). The other single ring-linker compounds generally had EC₅₀ values of <0.5 μM in this assay. The in vivo effects that may result from mitochondrial protein synthesis inhibition will require further investigation. It is worth noting that many antibiotics that work by inhibiting prokaryotic protein synthesis also inhibit mitochondrial protein synthesis (40). For example, the commonly used antibiotics such as doxycycline (COX-1 EC₅₀ = 6.6 μM) and linezolid (COX-1 EC₅₀ = 15 μM) are used at plasma levels that approximate or exceed the EC₅₀ determined in the COX-1 assay (41, 42). In the case of linezolid, toxicity due to mitochondrial inhibition can be observed during normal use of the drug, although this typically does not become serious until after 4 weeks of treatment (43). With anticipated treatment courses for cryptosporidiosis being relatively short (ideally no more than 3 days), it is likely that brief exposures to MetRS inhibitors would be well tolerated, although clearly this will require careful investigation. If mitochondrial inhibition is a problem, then additional effort to identify more selective compounds will be pursued in future work.

The compounds with the best in vivo activity (compounds 2093, 2114, and 2259) were also tested for inhibition of the hERG channels and CYP450 enzymes. The hERG inhibition assay screens for potential of the compound to promote dangerous cardiac dysrhythmias. In the presence of plasma protein (BSA), the percent inhibition was only ∼20% at concentrations of 30 μM, which is reassuring. The concentration that substantially inhibits hERG channels is likely to be several multiples above the peak concentrations that would occur during treatment. This same compound had little effect on CYP450 enzymes except for CYP2C8 (causing 62 to 87% inhibition at 10 μM). Common drugs that are metabolized by CYP2C8 include rosiglitazone (antidiabetic), montelukast (asthma), cerivastatin (statin), and amodiaquine (antimalarial). The coadministration of these MetRS inhibitors with the listed drugs could potentially lead to drug-drug interactions. Finally, Ames and in vitro micronucleus tests were done on compound 2093 and were negative. This provides reassurance that compound 2093 is not genotoxic.

In summary, the imidazopyridine compounds described here have potent activity against the C. parvum MetRS enzyme, as well as against cultures of C. parvum and C. hominis. Parasite persistence assays suggest the compounds have parasiticidal effects on the parasites. Most importantly, the MetRS inhibitors controlled C. parvum infection in two murine models without producing side effects. The active compounds demonstrated substantial plasma exposures, as well as fecal levels, but it is not entirely clear from these data which of these pharmacological parameters is most relevant. The MetRS inhibitors are capable of inhibiting the human mitochondrial MetRS enzyme as determined by reduced levels of a mitochondrial protein (COX-1) in cultured HepG2 cells; however, clinical toxicity may be unlikely if the duration of treatment is kept to short durations (e.g., <1 week). Future studies with compound 2093 and other MetRS inhibitors in the calf Cryptosporidium infection model, as well as in vivo toxicology studies, will help establish the potential for developing these compounds for the treatment of human cryptosporidiosis.

Support for this project was provided by PATH with funding from the UK government (grant 300341-111) and by the National Institutes of Health (grants AI097177 and HHSN272201700059C). The purified recombinant C. parvum MetRS protein (UniProt Q5CVN0) described here was provided by the Seattle Structural Genomics Center for Infectious Disease (www.SSGCID.org) supported by Federal Contract HHSN272201700059C from the National Institute of Allergy and Infectious Diseases, National Institutes of Health, Department of Health and Human Services.

We acknowledge Nora Molasky, Uyen Nguyen, and Omeed Faghih for performing mammalian cell cytotoxicity assays; Molly McCloskey and Grant Whitman for assisting with in vivo efficacy experiments; Zack Herbst for analytical chemistry on pharmacokinetic assays; and Hannah Udell for cloning and purifying the C. parvum MetRS enzyme. We acknowledge the contributions of project oversight by Eugenio de Hostos (PATH). We also acknowledge the valuable contributions made by collaborating scientists at Takeda (Yasushi Miyazaki) and CALIBR (Melissa Love and Case McNamara).