Discovery and Characterization of Clinical Candidate LXE408 as a Kinetoplastid-Selective Proteasome Inhibitor for the Treatment of Leishmaniases

Abstract

Visceral leishmaniasis is responsible for up to 30,000 deaths every year. Current treatments have shortcomings that include toxicity and variable efficacy across endemic regions. Previously, we reported the discovery of GNF6702, a selective inhibitor of the kinetoplastid proteasome, which cleared parasites in murine models of leishmaniasis, Chagas disease, and human African trypanosomiasis. Here, we describe the discovery and characterization of LXE408, a structurally related kinetoplastid-selective proteasome inhibitor currently in Phase 1 human clinical trials. Furthermore, we present high-resolution cryo-EM structures of the Leishmania tarentolae proteasome in complex with LXE408, which provides a compelling explanation for the noncompetitive mode of binding of this novel class of inhibitors of the kinetoplastid proteasome.

Introduction

Visceral leishmaniasis (VL) is an infectious disease causing 50,000–90,000 new cases and 30,000 deaths every year.¹ In the Indian subcontinent, the first-line treatment for this infection is a single infusion of AmBisome (liposomal amphotericin B). However, liposomal amphotericin B has limited efficacy in East Africa both as a monotherapy and in combination with other drugs;² therefore, in this endemic area the first line of treatment is a 17-day therapy with antimony stibogluconate and paromomycin combination.^3,4 Overall, existing parenteral treatments are difficult to administer, require hospitalization, and have dose-limiting toxicities.⁵ New drugs that are orally bioavailable, safe and can effect rapid cure are necessary to address needs of VL patients in developing economies.

Previously, we reported the identification of a novel pan-kinetoplastid growth inhibitor series, exemplified by GNF6702, and demonstrated that this triazolopyrimidine scaffold exerts its antiparasitic activity through selective inhibition of kinetoplastid proteasome chymotrypsin activity.⁶ Subsequently, a GSK/University of Dundee group reported the discovery of a structurally related scaffold, as well as its unique binding mode to Leishmania tarentolae proteasome, explaining the scaffold’s exceptional selectivity for the kinetoplastid compared to the mammalian proteasome.⁷ Here, we describe the optimization of GNF6702 that led to the selection of 1 (LXE408), a compound with remarkable efficacy in murine models of visceral and cutaneous leishmaniasis that is currently in Phase 1 human clinical trials. We also disclose the high-resolution cryo-EM costructure of 1 with the Leishmania tarentolae proteasome in the presence of the prototypical competitive proteasome inhibitor bortezomib which clarifies the noncompetitive nature of proteasome inhibition for this novel class of antileishmanial agents.⁶

Results and Discussion

The progression of GNF6702 (2) into clinical testing was hampered by solubility-limited oral absorption. While this limitation could have potentially been addressed by a specialized formulation (e.g., nanosuspension, solid dispersion), such an approach would likely increase the cost of treatment and complicate its use in developing economies. As a result, we set out to perform medicinal chemistry optimization of GNF6702 in order to identify a development candidate with improved solubility and improved oral exposure across preclinical species. Initial efforts, focusing on reducing logP and increasing the sp³ fraction to increase solubility, led to molecules with less potency or high clearance. A solubility analysis pointed to the high crystal packing energy of GNF6702 (melting point = 224 °C) as the key solubility-limiting issue.⁸ As the molecule was quite planar, we decided to introduce small substitutions at the pyridine 3-position (R) to force the pyridine out of plane with respect to the bicyclic core in an effort to reduce the crystal packing energy and thus improve solubility (Table 1).⁹

Table 1. In Vitro Data for Key Analogues.

compound no.	R group	L. donovani proteasome IC50 (μM)a	L. donovani EC50 (μM)b	HT sol. pH 6.8 (μM)c	ERd mouse/rat/dog/human	mp temp (°C)
1 (LXE408)	Me	0.04	0.04	17	<0.21/<0.18/<0.24/0.27	139
2 (GNF6702)	H	0.035	0.02	10	<0.21/<0.27/0.46/0.3	224
3	CHF2	0.06	0.11	21	0.51/<0.18/0.57/0.55	n.t.
4	Cl	0.1	0.05	7	<0.21/n.t. /<0.21/0.21	229
5	OMe	n.t.	0.07	7	0.4/n.t./0.56/<0.21	n.t.
6	CH2OH	0.07	0.5	n.t.e	n.t./n.t./n.t./n.t.	n.t.

^aHalf-maximum inhibition of chymotrypsin-like activity of purified L. donovani proteasome.

^bHalf-maximum inhibition of L. donovani amastigote proliferation within primary mouse macrophages.

^cSolubility determined in a high-throughput assay format.

^dMicrosomal extraction ratio. The assay measures the amount of parent compound remaining after 1 h incubation in microsomes.

^en.t. = not tested. IC₅₀ and EC₅₀ values correspond to means (n = 4 technical replicates).

In general, the compounds synthesized with different R groups maintained an inhibitory activity on purified L. donovani proteasome similar to the unsubstituted compound 2. Compounds 1 and 4 emerged as frontrunners, having good potency in the more relevant intracellular L. donovani growth inhibition assay and favorable microsomal stability across species. Finally, using the melting point temperature as an indicator of crystal packing energy, we selected compound 1 over compound 4 for further characterization due to its lower melting point temperature (melting point = 139 °C). A salt and form selection screen was conducted on compound 1 (22 different counterions and cocrystal formers in 4 solvents). The fumarate (cocrystal) was selected as the preferred candidate exhibiting appropriate properties in terms of stability, solubility, and excipient compatibility. The melting point of the thermodynamically stable polymorph fumarate cocrystal was 224 °C.

In a dissolution study, the fumarate cocrystal of 1 showed a significantly higher dissolution rate than the free base of 2 (see Figure 1). Furthermore, previous studies have shown that compounds with higher supersaturation stability in fasted state simulated intestinal fluid (FASSIF) display better absorption in the intestine.¹⁰ When compared to free base of compound 2, fumarate cocrystal of compound 1 was found to maintain a 5-fold higher supersaturation in FASSIF achieving a concentration of 73 μM at 16 min.¹¹ It should be noted that no suitable salt or cocrystal form of compound 2 with the right properties could be identified upon screening with various counterions. Overall, the data presented supported our decision to select compound 1 (LXE408) for further characterization.

Figure 1.

Time course of dissolution of 1 (LXE408, fumarate cocrystal) vs 2 (GNF6702).

To determine efficacy in vivo, compound 1 was administered for 8 days at three dose levels in a murine model of VL. In mice infected with L. donovani,¹² oral dosing (PO) with 1 effected a more pronounced reduction in liver parasite burden than miltefosine, the only oral antileishmanial drug available in clinical practice (Figure 2a). Dosing compound 1 at 1 mg/kg b.i.d. led to a 95% reduction of parasite burden in the liver, an efficacy equivalent to that effected by a 12 mg/kg q.d. regimen of miltefosine. This miltefosine regimen yielded plasma exposures in mice similar to the miltefosine clinical regimen. The in vivo efficacy of compound 1 plateaued at the 3 mg/kg b.i.d. dose with no further reduction of liver parasite burden at the 10 mg/kg b.i.d. dose. The plasma exposure of 1 increased dose-proportionally, and the free plasma concentration of 1 (mouse plasma protein binding = 94.4%) at the 3 and 10 mg/kg b.i.d. doses was above the EC₅₀ value of 1 on intramacrophage L. donovani amastigotes throughout the duration of treatment (Figure 2b). This observation is consistent with our previously published PK-PD model for this class of compounds, which indicated that the proteasome inhibitor efficacy is driven by free AUC and/or by time above C_min.⁶ Such a PK-PD profile suggests that sustained inhibition of the parasite proteasome, which results in intracellular accumulation of ubiquitylated proteins, is required for rapid clearance of L. donovani in vivo.

Figure 2.

a) Compound 1 post-treatment L. donovani liver burdens in a murine model of VL as assessed by qPCR. Circles: burdens in individual mice; thick horizontal lines: means of the treatment groups; red dotted line: limit of L. donovani detection by qPCR; circles below the red dotted line: mice with no detectable L. donovani parasites; RU: relative units. Data points below the limit of detection are “jittered” to show the number of animals in the group. b) Time profiles of mean free plasma concentration of compound 1; free compound 1 concentration values were predicted from measured total plasma concentration values collected on day 8 of treatment. Dashed red line corresponds to intramacrophage L. donovani EC₅₀ of 40 ± 8.1 nM. Circles: means ± SD; n = 5 mice; fraction unbound in mouse plasma = 0.056. For data points lacking error bars, standard deviations are smaller than circles the representing means.

Next, we tested the in vivo efficacy of 1 in a murine model of cutaneous leishmaniasis (CL). In BALB/c mice infected with L. major (a model of Old World leishmaniasis), oral dosing with 1 effected robust healing of parasite-induced skin lesions at the base of the tail. Administration of 1 for 10 days at a dose of 20 mg/kg b.i.d. produced a therapeutic effect comparable to that of liposomal amphotericin B, the most potent antileishmanial drug available today for CL therapy (Figure 3). Previously, we showed that compound 2 had superior antiparasitic activity in BALB/c footpad model of L. major infection when compared to miltefosine.⁶ Data presented here further extend these findings and establish 1 as a promising oral agent for treatment of cutaneous leishmaniasis.

Figure 3.

Efficacy of 1 in the BALB/c mouse model of Old World leishmaniasis. Data in the graph show progression of lesion sizes at the base of the tail of BALB/c mice infected with 1 × 10⁷L. major parasites after 10 days of treatment. Experimental time points represent the mean lesion sizes ± SEM (n = 6). Single-factor ANOVA (Dunnet’s test) was used to determine the statistical significance of differences in lesion sizes between vehicle and compound-treated groups (P < 0.05). Starting from day 4 post-treatment, lesion sizes in the vehicle control group were bigger and statistically differed from lesion sizes in compound-treated groups except for the compound 1 at 1 mg/kg b.i.d. dose.

Compound 1 was profiled in vitro for off-target activity on a panel of receptors including GPCRs, nuclear receptors, and kinases. For VMAT2, 61% inhibition was observed at 10 μM concentration; however, compound 1 had a low propensity to cross the blood brain barrier (brain/plasma AUC ratio = 0.03 in mice) suggesting minimal potential modulation of the VMAT2 receptor in the brain. Overall, compound 1 was not considered to pose a risk of off-target activity at clinically relevant concentrations (see the Supporting Information). In addition, the compound also showed no inhibition of the hERG channel (IC₅₀ > 30 μM) in a manual patch clamp assay indicating a low risk for QT prolongation. Finally, in vitro AMES and in vivo micronucleus (MNT) studies in rat did not indicate a risk of mutagenicity or clastogenicity, respectively.

Pharmacokinetic studies in selected preclinical species (mouse, dog, rat, and cynomolgus monkey) using the fumarate salt (Table 2) revealed that 1 had low clearance (CL, 2–17% of hepatic blood flow) and low steady-state volume of distribution values (V_ss, smaller or comparable to the volume of total body water) in mouse, rat, and monkey following intravenous (IV) compound administration. However, the CL and V_ss values in dog were in the moderate range (CL: 37% of hepatic blood flow and V_ss larger than the total volume of body water). The plasma terminal half-life of 1 was in the range of 3–10 h across species following IV administration. Oral bioavailability was 27–67% across the tested species when using conventional suspension formulations.

Table 2. Pharmacokinetic Profile of Compound 1 in Selected Preclinical Species.

PK parameters	mousea	ratb	dogc	monkeyd
CL (mL/min/kg)	2.3	2.1	11.6	0.9
Vss (L/kg)	0.63	0.53	2.1	0.5
T1/2 IV (h)	3.3	3.8	3.8	9.7
MRT IV (h)	4.6	4.3	3.3	9.7
DN AUC PO (h·μM)	7.5	12	1.5	12
DN Cmax PO (μM)	0.8	1.0	0.3	1.0
F (%)	46	67	44	27

^aBalb/C mice, 5 mg/kg IV and 20 mg/kg PO doses.

^bMale Sprague–Dawley rat, 3 mg/kg IV and 10 mg/kg PO doses.

^cMale beagle dog, 0.3 mg/kg IV and 1.0 mg/kg PO doses.

^dMale cynomolgus monkey, 0.3 mg/kg IV and 10 mg/kg PO doses. For IV dosing, 1 was formulated in 75% PEG300/25% D5W vehicle; for PO dosing, 1 was formulated in 0.5% methylcellulose, 0.2% Tween 20 vehicle; CL: plasma clearance after administration of a single IV bolus dose; V_ss: volume of distribution at steady-state; DN AUC PO: dose-normalized area under the plasma concentration–time curve after administration of a single oral dose; DN C_max PO: dose-normalized maximum plasma concentration after administration of a single oral dose; F (%): oral bioavailability after administration of a single oral dose.

In order to select the preclinical species for toxicology studies that would reflect most closely the metabolism of 1 in human, we identified and quantified the most abundant metabolites in in vitro hepatocytes and in vivo. In vitro, 1 was mainly metabolized to monohydroxylated metabolites 6–8 (see Figure 4). Metabolite 6 showed a modest activity of 0.5 μM on parasites, whereas the other two metabolites had an activity of greater than 5 μM. No unique metabolites were detected in human hepatocytes. While the metabolism profile of 1 was qualitatively similar across the tested species, we observed species differences in the extent of formation of individual metabolites. The predominant metabolite in dog hepatocytes was 8 with minor formation of the other two metabolites (6 and 7). In contrast, in mouse, rat, monkey, and human hepatocytes, the predominant metabolite was 7 with minor formation of the other two metabolites (6 and 8).

Figure 4.

Compound 1 main oxidative metabolites.

The in vivo metabolic profiles of 1 in mouse, rat, dog, and monkey plasma after oral administration were consistent with the in vitro hepatocytes data (Table 3). After a single oral dose, the exposure of metabolite 8 was higher than that of 1 in dog plasma (120%), while it was minor (0.2–5%) in plasma of all the other species (see Table 4). Levels of 6 and 7 were in moderate or low range in dog plasma (≤10% of parent). In mouse, rat, and monkey plasma, 6 was less than 10% relative to the equivalent exposure of 1, and 7 was in the range of 5–17% of parent exposure. It is worth noting that the aniline resulting from amide hydrolysis was not detected in vivo. Using a panel of dog recombinant CYP enzymes, we performed P450 phenotyping studies that demonstrated that 1 was metabolized to 8 predominantly by CYP2D15 in dog. Importantly, compound 1 was not a substrate of human CYP2D6, the human ortholog of dog CYP2D15. Human recombinant CYP phenotyping experiments showed that CYP3A4, CYP1A1, 1A2, and 2J2 were the major enzymes responsible for the formation of metabolite 8. These data suggested that the metabolism in dog was not predictive of human metabolism, and therefore, monkey was selected as the nonrodent species for toxicological studies.

Table 3. Distribution of Metabolites in Various Preclinical Species^a.

	abundancea
	mouse	rat	dog	monkey	human
1	+++	+++	+++	+++	+++
6	+	+	+	+	+
7	+	+	+	+	+
8	n.d.	+	++	+	+

^aSemiquantitative metabolite profiles were obtained after incubation of 10 μM of 1 for 4 h and analysis using UV chromatographic data (250 nm). All integrable peaks were normalized to the percent area of the largest observed peak. +++ > 75%; ++ ≤ 75%>25%; + ≤25%.

Table 4. Percentage of Metabolite Exposure Relative to 1 in Preclinical Species Following Oral Administration at 10 mg/kg.

	metabolites (%)a,b
species	6	7	8
mouse	1.5	4.5	0.2
rat	1.5	17	1.5
dog	9.1	2.3	120
monkey	1.7	11	5.1

^aAUC of metabolites was dose-normalized before relative amounts were calculated.

^bThe concentrations were analyzed from experiments described in Table 2.

To further characterize compound 1, we wanted to establish its mode of binding to the proteasome.

Previously published mechanism of action studies indicated that 2 is an inhibitor of the kinetoplastid proteasome that selectively inhibits the chymotrypsin-like activity of the β5 proteasome subunit (PSMB5). In a biochemical assay, 2 inhibited the chymotrypsin-like activity in a noncompetitive manner when the fluorogenic peptide Suc-LLVY-[Rh110]-[Dpro] was used as the substrate. This observation, together with the identified resistant mutations in the β4 proteasome subunit (PSMB4), pointed to a potential binding site for 2 at the interface of the PSMB4 and PSMB5 subunits.⁶ However, the exact binding site of 2 and how the compound binding mode relates to its noncompetitive mode of inhibition remained unanswered.⁷ To this end, we elucidated the atomic structure of 1 in complex with the 20S proteasome from L. tarentolae using cryo-EM (Figure 6a). The resultant 3.4 Å resolution structure revealed a clearly defined density for compound 1 (Figure 6b) with 1 situated within the active site pocket of PSMB5 partly formed by the side chains of Ile29 and Phe24 of the PSMB4 subunit (Figure 6c).¹³ There, the dimethyl-oxazole unit of 1 packs tightly with the two hydrophobic residues. The conformation is analogous to the one recently published by Wyllie et al.⁷ and is in agreement with binding profiles from parasite resistant mutations. First, the β5 Y113F mutation of T. brucei proteasome leads to loss of activity for compound 1 (Supporting Information; 113 refers to the amino acid position in the mature β5 subunit), and second, the F24L mutation of the L. tarentolae proteasome exhibits a significant upshift in the IC₅₀ value in response to the inhibitor (Figure 5). This is in contrast to the binding curve for bortezomib (Figure 5) which shows no reduction in activity in response to the resistant mutation. Bortezomib is a prototypical peptidomimetic inhibitor of the proteasome, which inhibits proteasome chymotrypsin-like activity competitively with peptide substrates such as Suc-LLVY-[Rh110]-[Dpro]. This lack of resistance is in complete agreement with a large number of structural and biochemical studies of bortezomib in homologous proteasomes.¹⁴

Figure 6.

Binding of 1 to L. tarentolae proteasome. a) Positions of the two compound 1 molecules in the β5 subunit of proteasome, related β5 domains colored cyan, β4 subunits colored green. Compound 1 carbon atoms colored as yellow spheres. b) Electron density of the 3.4 Å structure around compound 1 (at 3 sigma deviations above the mean. EMDB and PDB code: EMD-10472, PDB ID 6TCZ. c) Key interactions around compound 1, with resistance mutants in β4 subunit colored as orange sticks.

Figure 5.

F24L mutation in proteasome β4 subunit confers selective resistance to 1. Inhibition of chymotrypsin-like activity of purified wild-type (PSMB4^WT) and PSMB4^F24LL. tarentolae proteasome by 1 and bortezomib; IC₅₀ values of chymotrypsin-like activity are listed inside the plot. Data shown represents the means from one of the representative experiments (n = 3 technical replicates). RU (relative units) corresponds to the percentage of chymotrypsin-like activity relative to the DMSO control.

However, when the structure of the canonical bortezomib binding mode is superimposed with the structure of the binary complex with 1, steric clashes indicate that 1 and bortezomib cannot bind to the PSMB5 pocket simultaneously without a conformational change in the pocket for one or both of the compounds. To understand this at an atomic level, a cryo-EM structure of the ternary complex with L. tarentolae proteasome in the presence of 1 and bortezomib was solved. The resulting proteasome structure, solved at 3.2 Å resolution, clearly revealed that both 1 and bortezomib can occupy the pocket as a ternary complex with the proteasome. The binding mode of 1 remained unchanged when compared to the binary complex between the compound and the proteasome. In contrast, bortezomib adopted a novel conformation with its phenyl group rotated ∼60 degrees around the dihedral angle connection formed by the phenyl with the bortezomib backbone when compared to the previously published crystal structures of the human proteasome in complex with bortezomib (Figure 7a and 7b).¹³ Overall, these findings provide atomic level resolution details of the binding of 1 to the kinetoplastid proteasome and explain its noncompetitive mode of chymotrypsin-like activity inhibition.

Figure 7.

a) Final density map of the ternary complex of compound 1 and bortezomib bound to the β5 pocket. The map is displayed as a blue isomesh contoured at 3.0 σ above the mean, and for clarity it is displayed only around the ligands at a radius of 2.0 Å. The β5 chain is colored green in a cartoon representation; 1 and bortezomib are represented as sticks with carbons colored as yellow and beige, respectively. b) Rotation of bortezomib when competing with 1, bortezomib in beige, 1 in yellow, relative to bortezomib. Leishmania tarentolae proteasome 20S subunit complexed with 1 and bortezomib accession code(s): EMD-10463, PDB ID: 6TD5.

Chemical Synthesis of Compound 1

Scheme 1 outlines the research synthesis of compound 1 that was achieved in 8 steps with an overall 8% yield. The synthesis was initiated from commercially available 2-fluoro-5-nitrobenzoic acid 9, which upon treatment with thionyl chloride was converted to acid chloride. The excess thionyl chloride was removed under reduced pressure to be used for next step. The crude acid chloride was resuspended in toluene, and the intermediate was treated with aminoguanidine carbonate for 12 h, which yielded 10 as a solid. Amino guanidine intermediate 10 was stirred in water at 100 °C for 8 h to yield the amino-triazole 11 in 51% yield. Triazole 11 was condensed with bromomalonaldehyde, and the resulting nitro compound was further reduced with Raney nickel in the presence of zinc iodide to afford the key intermediate 12. The resulting amide from coupling of aromatic amine 12 with 2,4-dimethyl oxazole carboxylic acid was further functionalized to boronic ester using a Miyaura Borylation reaction.¹⁵ Finally, Suzuki coupling between the boronic acid and 2-bromomethylpyridine gave the desired compound 1.

Scheme 1. Synthesis of Compound 1.

Reagents and conditions. (a) SOCl₂, 80 °C, 4 h, 98%; (b) aminoguanidine carbonate, toluene, room temp 12 h, 97%; (c) H₂O, AcOH-carboxylic acid, HATU, DIEA, DMF, 0.5 h, 53%; (d) bromomalonaldehyde, AcOH, 100 °C, 43% (e) raney nickel, ZnI₂, THF, H₂ @ 50 psi, 2 h, 77% (f) 2,4-dimethyloxazole-5-carboxylic acid, HATU, DIEA, DMF, 0.5 h, 53%; (g) 4,4,4′,4′,5,5,5′,5′-octamethyl-2,2′-bi (1,3,2-dioxaborolane), Pd(dppf)Cl₂, KOAc, dioxane, 100 °C, 16 h, 24%; and (h) 2-bromo-3-methylpyridine, Pd(dppf)Cl₂, 1M, aq Na₂CO₃, 80 °C, 16 h, 39%.

Conclusions

The current target product profile for a new antileishmanial drug is to be safe, effective, and oral with a short course of treatment (<11 days).¹⁶ Optimization of GNF6702 successfully led to compound 1 (LXE408) which, when dosed orally, demonstrates excellent efficacy in mouse disease models and has an appropriate safety profile in preclinical species. Applying the PK-PD principles, we expect that a dose of 85–190 mg per day will effect cure in VL patients. Currently, phase I studies in healthy human volunteers are ongoing to determine safety and tolerability. The compound will most likely be dosed in combination with other agents once the safety and efficacy is established in humans as a single agent. Cryo-EM efforts led to a detailed understanding of the mechanism of inhibition of chymotrypsin activity by 1 and elucidated a novel mode of binding. The discovery of the noncompetitive specific β5 proteasome inhibition and further understanding of the binding mode of 1 could prove to be transformative in the fields of infectious diseases and oncology where more subunit-selective proteasome inhibitors are being sought.^17,18

Experimental Section

Materials and Methods

Commercially available starting materials were used as supplied without further purification. Reactions were carried out by using dry organic solvents (DCM, ACN, DMF, etc.) unless otherwise noted. Reactions were monitored using thin layer chromatography and an Agilent Technologies 1200 series 6140 Quadrupole LC-MS with UV detection at 254 nm in electrospray ionization (ESI) mode. For LC-MS, all retention times reported are at 254 nm UV channel unless otherwise noted. All NMR spectra were recorded on a Bruker AVANCE-400 spectrometer operating at a frequency of 400.13 MHz for ¹H and 100.61 MHz for ¹³C equipped with a 5 mm QNP cryoprobe with Z-gradient. Chemical shifts for ¹H and ¹³C spectra were referenced to residual solvent. MS were obtained on an Agilent Technologies 1200 series 6140 quadrupole LC-MS in electrospray ionization (ESI) mode. HRMS-ESI data were recorded using an Agilent 6520 Accurate-Mass Q-TOF LC-MS system with HPLC-Chip Cube interface and an Agilent 1200 HPLC. All final compounds were isolated analytically pure, >99% purity by HPLC unless otherwise indicated. More information can be found in the Supporting Information.

2-(2-Fluoro-5-nitrobenzoyl)hydrazine-1-carboximidamide (10)

A solution of 2-fluoro-5-nitrobenzoic acid (50 g, 270 mmol) in thionyl chloride (100 mL) was heated to 80 °C and stirred for 4 h. The mixture was allowed to cool down to room temperature, and the solvent was removed to give acid chloride intermediate (54 g, 98% yield).

To a solution of aminoguanidine carbonate (36.2 g, 266 mmol) in dry toluene (300 mL) cooled to 0 °C was added the above acid chloride (54 g, 266 mmol) over 30 min. The mixture was stirred at room temperature for 12 h. The formed precipitate was removed by filtration, and the residue was treated with H₂O (400 mL) and made alkaline with sodium carbonate. The solid was collected and recrystallized from water to obtain compound 10 (62 g, 97% yield). LCMS: [ESI] m/z = 241.1 [M + H]⁺. The synthesis of the compound has been described previously and is commerically available.

5-(2-Fluoro-5-nitrophenyl)-4H-1,2,4-triazol-3-amine (11)

A solution of compound 10 (62 g, 257 mmol) in H₂O (800 mL) was stirred for 8 h at 100 °C. After cooling, the solid obtained was filtered, and the cake was washed with H₂O (100 mL) and THF (100 mL) and dried to give 11 (34 g, 51% yield). ¹H NMR (400 MHz, DMSO) 12.42 (s, 1H), 8.74 (dd, J = 6.27 Hz, 3.0 Hz, 1H), 8.26 (dt, J = 7.0 Hz, 3.4 Hz, 1H), 7.57 (t, J = 9.5 Hz, 1H), 6.29 (s, 2H). ¹³C NMR (101 MHz, DMSO) δ 164.34, 161.71, 157.72, 144.42, 144.39, 125.76, 125.66, 125.25, 125.20, 118.85, 118.61. HRMS found [M + H]⁺ = 224.0578, C₈H₇O₂N₅F predicted [M + H]⁺ = 224.0578.

3-(6-Bromo-[1,2,4]triazolo[1,5-a]pyrimidin-2-yl)-4-fluoroaniline (12)

To a solution of compound 11 (10 g, 44.8 mmol) in acetic acid (50 mL) was added 2-bromomalonaldehyde (8.12 g, 53.8 mmol) at room temperature. The mixture was heated to 100 °C and stirred for 4 h. The mixture was allowed to cool to room temperature before adding water (100 mL) and filtering. The filter cake was then washed with THF and dried under vacuum to give compound 6-bromo-2-(2-fluoro-5-nitrophenyl)-[1,2,4]triazolo[1,5-a]pyrimidine (6.5 g, 43% yield). ¹H NMR (400 MHz, DMSO) 10.04 (s, 1H), 9.08 (s, 1H), 8.96–9.04 (m, 1H), 8.47 (dt, J = 9.0 Hz, 3.4 Hz, 1H), 7.76 (t, J = 9.5 Hz, 1H). ¹³C NMR (101 MHz, DMSO) δ 165.01, 162.36, 160.38, 157.47, 154.11, 144.68, 138.27, 138.09, 128.47, 128.36, 126.51, 126.46, 119.54, 119.36, 119.30, 107.08. HRMS found [M + H]⁺ = 337.9688, C₁₁H₆O₂N₅BrF predicted [M + H]⁺ = 337.9683.

To a solution of compound 6-bromo-2-(2-fluoro-5-nitrophenyl)-[1,2,4]triazolo[1,5-a]pyrimidine (6 g, 17.7 mmol) in THF (150 mL) was added Raney Nickel (7 g) and ZnI₂ (2.26 g, 7.1 mmol), the suspension was degassed under vacuum, and the mixture was stirred under H₂ (50 psi) at 25 °C for 2 h. The mixture was filtered, and the solvents were removed to give the crude product. The crude product was washed with MeOH (50 mL × 2) and dried under vacuum to give compound 12 (4.2 g, 77% yield).¹H NMR (400 MHz, DMSO): 9.93 (d, J = 2.3 Hz, 1H), 8.98 (d, J = 2.0 Hz, 1H), 7.39–7.45 (m, 1H), 6.99–7.10 (m, 1H), 6.67–6.76 (m, 1H), 5.76 (s, 1H), 5.25 (brs, 2H). ¹³C NMR (101 MHz, DMSO) δ 163.08, 156.52, 154.05, 153.74, 151.32, 145.84, 137.87, 118.03, 117.91, 117.80, 117.73, 117.56, 117.33, 114.86, 106.16. HRMS found [M + H]⁺ = 307.9941, C₁₁H₈O₂N₅BrF predicted [M + H]⁺ = 307.9942.

N-(4-Fluoro-3-(6-(3-methylpyridin-2-yl)-[1,2,4]triazolo[1,5-a]pyrimidin-2-yl)phenyl)-2,4-dimethyloxazole-5-carboxamide (1)

Step F: To a solution of 2,4-dimethyloxazole-5-carboxylic acid (1.92 g, 13.6 mmol) in DMF (50 mL) was added HATU (6.2 g, 16.32 mmol) and DIEA (3.5 g, 27.2 mmol) at room temperature. The mixture was stirred for 30 min, and the compound 12 (4.2 g, 13.6 mmol) was added. The mixture was stirred for 3 h, and water (100 mL) was added. The mixture was filtered, and the filter cake was washed with H₂O (50 mL × 2) and THF (50 mL × 2) and dried to give N-(3-(6-bromo-[1,2,4]triazolo[1,5-a]pyrimidin-2-yl)-4-fluorophenyl)-2,4-dimethyloxazole-5-carboxamide (3.1 g, 53%). ¹H NMR (400 MHz, DMSO) 10.41 (s, 1 H), 9.97 (s, 1H), 9.02 (brs, 1H), 8.75 (s, 1H), 7.93 (s, 1H), 7.40 (t, J = 9.66 Hz, 1H), 2.50 (s, 3H), 2.39 (s, 3H). ¹³C NMR (101 MHz, DMSO) δ 162.29, 161.49, 156.86, 156.78, 154.16, 143.51, 139.17, 138.05, 135.52, 124.95, 122.80, 117.67, 117.44, 106.48, 14.29, 13.26. HRMS found [M + H]⁺ = 431.0259, C₁₇H₁₃O₂N₆BrF predicted [M + H]⁺= 431.0262.

Step G: To a solution of compound N-(3-(6-bromo-[1,2,4]triazolo[1,5-a]pyrimidin-2-yl)-4-fluorophenyl)-2,4-dimethyloxazole-5-carboxamide (3.0 g, 7.0 mmol) in dioxane (60 mL) was added 4,4,4′,4′,5,5,5′,5′-octamethyl-2,2′-bi(1,3,2-dioxaborolane) (7.1 g, 28 mmol), [1,1′-bis(diphenylphosphino)ferrocene]dichloropalladium(II) in complex with dichloromethane (0.5 g, 0.7 mmol), and KOAc (2.0 g, 21 mmol) under N₂. This mixture was heated to 100 °C for 16 h. The solvent was removed, and the compound was redissolved in THF and filtered through silica gel. The filtrate was concentrated and triturated with MTBE to give compound (2-(5-(2,4-dimethyloxazole-5-carboxamido)-2-fluorophenyl)-[1,2,4]triazolo[1,5-a]pyrimidin-6-yl)boronic acid (0.67 g, yield, 24%). ¹H NMR: DMSO 400 MHz: 2.40 (s, 3 H), 2.51 (s, 3 H), 7.38–7.43 (m, 1 H), 7.92–7.96 (m, 1 H), 8.72–8.79 (m, 2 H), 9.08–9.09 (m, 1 H), 9.45–9.48 (m, 1 H), 10.41 (s, 1 H). ¹³C NMR (101 MHz, DMSO) δ 162.10, 161.49, 160.65, 158.01, 156.83, 155.93, 155.49, 143.47, 142.17, 139.18, 135.43, 124.85, 122.93, 118.51, 118.39, 117.60, 117.38, 14.28, 13.26. LCMS: [ESI] m/z 397.1 [M – 82 + 1]⁺. HRMS found [M – 82 + 1]⁺ = 397.1230, C₁₇H₁₅O₄N₆BF predicted [M – 82 + 1]⁺ = 397.1226.

N-(4-Fluoro-3-(6-(3-methylpyridin-2-yl)-[1,2,4]triazolo[1,5-a]pyrimidin-2-yl)phenyl)-2,4-dimethyloxazole-5-carboxamide (1)

To a solution of (2-(5-(2,4-dimethyloxazole-5-carboxamido)-2-fluorophenyl)-[1,2,4]triazolo[1,5-a]pyrimidin-6-yl)boronic acid (100 mg, 0.21 mmol) and 2-bromo-3-methylpyridine (54 mg, 0.31 mmol) in 1,4-dioxane (10 mL) was added 1 N aqueous sodium carbonate (1.1 mL, 1.1 mmol) and [1,1′-bis(diphenylphosphino)ferrocene]dichloropalladium(II) in complex with dichloromethane (8.5 mg, 0.011 mmol) at room temperature. The reaction was purged with N₂ for 1 min and stirred at 80 °C for 16 h. The reaction mixture was diluted with ethyl acetate and washed with saturated NaHCO₃ and brine. The organic layer was dried over magnesium sulfate, filtered, and reduced to dryness. The crude product was purified by reverse phase HPLC to afford compound 1 (36 mg) as a white solid. ¹H NMR (400 MHz, MeOD) 9.57 (d, J = 2.4 Hz, 1H), 9.14 (d, J = 2.3 Hz, 1H), 8.67 (dd, J = 5.1 Hz, 1.6 Hz, 1H), 8.58 (dd, J = 6.4 Hz, 2.7 Hz, 1H), 8.09 (dd, J = 7.8 Hz, 1.6 Hz, 1H), 7.91 (ddd, J = 8.9 Hz, 4.3 Hz, 2.8 Hz, 1H), 7.63 (dd, J = 7.9 Hz, 5.1 Hz, 1H), 7.34 (dd, J = 10.4 Hz, 9.0 Hz, 1H), 2.56 (s, 6H), 2.48 (s, 3H). ¹³C NMR (101 MHz, DMSO) δ 161.92 (d, J = 5.1 Hz), 161.03, 156.79, 156.34, 156.25 (d, J = 252.9 Hz), 154.25, 150.73, 147.4, 143.03, 139.26, 138.69, 136.36, 135.02 (d, J = 2.9 Hz), 132.38, 124.31 (d, J = 8.4 Hz), 123.73, 123.62, 122.38, 117.94 (d, J = 12.1 Hz), 117.07 (d, J = 22.7 Hz), 19.34, 13.85, 12.82. LCMS: [ESI] m/z = 444.4 [M + H]⁺). HRMS found [M + H]⁺, 444.1577, C₂₃H₁₉O₂N₇F predicted [M + H]⁺ = 444.1579.

Glossary

Abbreviations

PSMB

proteasome subunit beta type

cryo-EM

cryogenic electron microscopy

VL

visceral leishmaniasis

ER

extraction ratio

HT

high throughput

FASSIF

fasted state simulated intestinal fluid

qPCR

quantitative polymerase chain reaction

EC₅₀

half maximal effective concentration

IC₅₀

half-maximum inhibitory concentration

PO

oral administration

IV

intravenous

b.i.d.

twice a day

q.d.

once daily (“quaque die”)

AUC

area under the curve

DN AUC

dose normalized area under the curve

MRT

mean residence time

PK-PD

pharmacokinetic-pharmacodynamics

SD

standard deviation

SEM

standard error of the mean

MNT

micronucleus test

GPCR

G-protein coupled receptor

VMAT2

vesicular monoamine transporter 2

CYP

cytochrome P

HRMS

high-resolution mass spectrometry

LCMS

liquid chromatography mass spectrometry

ESI

electrospray ionization

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.jmedchem.0c00499.

• Molecular formula strings (CSV)

• Experimental procedures and analytical data for synthesis of all compounds and detailed characterization of compound 1 and protocols for in vitro assays and in vivo studies strings (PDF)

Accession Codes

The cryo-EM structure of binding mode of of 1 to L. tarentolae proteasome (accession codes: EMD-10472, PDB ID 6TCZ) and Leishmania tarentolae proteasome 20S subunit complexed with 1 and bortezomib (accession codes: EMD-10463, PDB ID: 6TD5) have been deposited in EMDB and PDB databases. The authors will release the atomic coordinates upon article publication.

Author Present Address

^○ For R.J.S: National Institutes of Health, Office of Biodefense, Research Resources and Translational Research, 5601 Fishers Lane, Bethesda, Maryland 20892, United States.

Author Present Address

^± For V.Y., M.G.: Strateos Inc., San Diego, California 92121, United States.

Author Present Address

^§ For I.L.: Gossamer Bio, San Diego, California 92121, United States.

Author Present Address

^∥ For Y.X.: EOC Pharma, Shanghai, China.

Author Present Address

^◇ For A.B.: ashley.m.berman@gmail.com.

Author Present Address

^◆ For L.C.D.: ldavis2199@gmail.com.

Author Present Address

^¶ For S.K.: Ferring Research Institute Inc., San Diego, California 92121, United States.

Author Present Address

^● For K.J.: Neurocrine Biosciences, Inc., San Diego, California 92130, United States.

Author Present Address

^Δ For T.T.: Tovetuntland@outlook.com.

Author Present Address

^£ For R.J.G.: Inception Sciences, Inc., San Diego, California 92121, United States.

This work was supported in part by grants from the Wellcome Trust (091038/Z/09/Z to R.J.G. and F.S.).

The authors declare the following competing financial interest(s): The following authors are employees or former employees of GNF or NITD or NIBR: A.S.N., A.B., C.B., S.H., A.H., B.B., Y.H.L., M.-Y.G., F.L., C.J.N.M., X.L., V.Y., J.S., I.L., Y.X., D.C., M.G., A.B., Y.-L.C., J.J., L.C.D., X.L., J.B., S.K., F.K.E., A.L., T.G., M.S., W.R., K.J., P.J.R., S.P.S.R., T.T., G.S., R.J.G., F.S., C.W., and V.M.