Structure-Guided Design and Synthesis of a Pyridazinone Series of Trypanosoma cruzi Proteasome Inhibitors

There is an urgent need for new treatments for Chagas disease, a parasitic infection which mostly impacts South and Central America. We previously reported on the discovery of GSK3494245/DDD01305143, a preclinical candidate for visceral leishmaniasis which acted through inhibition of the Leishmania proteasome. A related analogue, active against Trypanosoma cruzi, showed suboptimal efficacy in an animal model of Chagas disease, so alternative proteasome inhibitors were investigated. Screening a library of phenotypically active analogues against the T. cruzi proteasome identified an active, selective pyridazinone, the development of which is described herein. We obtained a cryo-EM co-structure of proteasome and a key inhibitor and used this to drive optimization of the compounds. Alongside this, optimization of the absorption, distribution, metabolism, and excretion (ADME) properties afforded a suitable compound for mouse efficacy studies. The outcome of these studies is discussed, alongside future plans to further understand the series and its potential to deliver a new treatment for Chagas disease.


■ INTRODUCTION
The kinetoplastids, a group of related parasites, are responsible for diseases which cause an enormous health and economic burden on countries in tropical regions of the world. 1 One of these, Chagas disease, which is caused by infection with Trypanosoma cruzi (T. cruzi), is mainly found in South America, Central America, and Mexico, with estimates suggesting there are 6−7 million infected people worldwide, resulting in around 10,000−14,000 deaths per annum. 2−5 Chagas disease is primarily transmitted via an insect vector, the triatomines (kissing or vampire bugs), although other routes of transmission do occur. The disease passes through a series of distinct phases: after the initial acute infection, patients enter the indeterminate phase, which is characterized by low levels of parasites and lack of overt symptoms. A subset of patients develops symptomatic disease with serious health conditions such as cardiomyopathy and digestive disorders.
There are currently only two approved treatments for Chagas disease, benznidazole (BNZ) and nifurtimox, 6,7 whose use is hampered by issues including limited efficacy and side effects which can lead to early treatment discontinuation. Recent clinical trials, such as the BENDITA trial (NCT03378661), 8 suggest improved dosing regimens for BNZ might be beneficial, but the pipeline of alternative treatments is still very limited, incorporating fexinidazole and the oxaborole compounds DNDi-6148 9,10 and AN15368. 11 There is therefore an urgent need for new treatments, particularly ones with differentiated modes of action compared to those already in the clinic and those which have been shown to be ineffective (such as CYP51 inhibitors). 12−15 A potential start-point for new therapies would be compound series that target the related parasites Trypanosoma brucei and Leishmania, whose mechanism of action has the potential to be applied to Chagas disease. From a collaboration between the Drug Discovery Unit at the University of Dundee and GSK's Global Health Research Centre in Tres Cantos, we recently disclosed the clinical candidate GSK3494245/DDD01305143 1 for visceral leishmaniasis (VL) (Figure 1). 16,17 While the series was developed phenotypically, it was subsequently demonstrated to act through inhibition of the β5 (chymotrypsin) proteolytic activity of the parasite proteasome, with a cryo-EM structure of the proteasome−inhibitor complex being generated. The proteasome is a multi-subunit protease that is conserved between higher eukaryotes and protozoan parasites and is a key component of the ubiquitin-proteasome system which is essential for the maintenance of intracellular protein homeostasis. 18 The core particle of the proteasome consists of 28 subunits organized into four rings of seven subunits and harbors the three proteolytic activities: caspase-like (β1 subunit, preferential cleavage after acidic amino acids), trypsin-like (β2 subunit, preferential cleavage C-terminal to positively charged amino acids), and chymotrypsin-like (β5 subunit, preferential cleavage after hydrophobic amino acids). 19,20 Due to its essential role in protein homeostasis, the proteasome is expected to be a good drug target across protozoan infectious diseases. 21−23 Specifically for T. cruzi, in vivo chemical validation of the proteasome as a suitable drug target was obtained with the Novartis compounds GNF6702 2 24 and LXE408 3, 25 two compounds related to 1. This series has pan-kinetoplastid activity and 3 is currently progressing to clinical trials for VL ( Figure 1).
Herein, we disclose our efforts to repurpose the series, which led to 1 into Chagas disease, and a subsequent screening campaign to identify novel chemical start-points. We report the use of cryo-EM to aid design, the development of compounds suitable for dosing in a chronic mouse model of Chagas disease, and the results of these efficacy studies.

■ RESULTS AND DISCUSSION
Repurposing of Antileishmanial Proteasome Inhibitors for Chagas Disease. From the previously reported series of Leishmania proteasome inhibitors, 16,17 one of the compounds, 4, was profiled for its potential repurposing for Chagas disease. As shown in Table 1, 4 was a potent, selective inhibitor of the T. cruzi proteasome, 26 and this translated into potency in the intracellular T. cruzi assay 27 with a pEC 50 of 6.9 and no activity against the host VERO cells. 4 was tested in an in vitro washout assay, 28 and we were encouraged to see that after 8 days of treatment 4, at 50-fold EC 50 , performed better than BNZ, with parasites relapsing on day 34, compared to day 21 for BNZ. 4 had previously been run in a mouse efficacy model of VL, dosed at 50 mg/kg and plotting exposure obtained from this study against the T. cruzi EC 99 ( Figure 2) showed that exposure in blood was well above the T. cruzi EC 99 for >8 h, even when correcting for plasma protein binding. Although these were VL-  infected mice, we have not generally seen marked differences in drug exposure between uninfected and Leishmania/T. cruziinfected mice (unpublished data). Based on its in vitro profile, together with the pharmacokinetic profile obtained from the VL efficacy study, 4 was progressed into a chronic mouse efficacy model of Chagas disease. 30 In this case, after treating the mice for 5 days at 50 mg/kg bid, there was no visible effect on parasite levels even though exposure was consistent with the VL efficacy study (BNZ dosed at 100 mg/kg reduced parasite levels below detectable limits).
Based on the promising results in the washout study, we were keen to understand the lack of effects in the in vivo study, where differences between parasite and compound distribution could be an important factor, with the parasites being widely distributed to sites including adipose tissue, ovaries, and the CNS. 31,32 Monitoring both parasite and compound distribution in vivo would be challenging so we elected to utilize CNS exposure as a surrogate for wider compound distribution to the other conserved sites where the parasite resides. The brain− blood ratio for 4 was shown to be 0.07, suggesting limited CNS penetration, likely due to 4 being a strong Pgp substrate with an efflux ratio (ER) of 22 in an MDR1-MDCK assay. Indeed, predicted brain exposure of 4 was well below T. cruzi EC 99 (Figure 3).
A large number of analogues of 4 were screened in the MDR1-MDCK assay, but no potent, low-efflux compounds were identified (data not shown). Because of this, we decided to utilize our knowledge of the mechanism of action of the series and our capability to generate cryo-EM structures to initiate a hit-discovery campaign against the T. cruzi proteasome.
Identification of T. cruzi Proteasome Inhibitors. Approximately 26,000 compounds from the GSK collection with activity against T. cruzi 33 were screened in a T. cruzi proteasome assay with actives also tested in a human  . Predicted free exposure of 4 in the CNS, based on the blood−brain ratio of 0.07, brain tissue free fraction of 4.3%, and the blood exposure from the chronic mouse efficacy model of Chagas disease. Solid blue line = day 1 predicted free brain concentration; solid red line = day 5 predicted free brain concentration; solid black line = T. cruzi EC 99 . proteasome assay 26 in order to identify hits that were potent against the parasite proteasome and selective against the human proteasome, as well as having antiparasitic activity. Although there was no guarantee that intracellular activity was solely driven by proteasome inhibition, this could be monitored as a hit series was developed. Any hits would also be screened in the MDR1-MDCK assay, with the aim of identifying a hit with markedly reduced Pgp substrate interaction compared to 4.
This screening cascade led to the identification of 5, a functionalized pyridazinone with our desired potency/selectivity profile and reasonable LE/LLE (0.29, 2.5), albeit with an ∼10-fold drop-off in potency from T. cruzi proteasome inhibition to antiparasitic activity. Further profiling showed 5 to have good aqueous solubility, reasonable lipophilicity, and high permeability in the MDR1-MCDK assay ( Table 2). The ER was 6, which we considered a reasonable start-point for optimization, especially compared to 4 with an ER of 22.
Ideally, we aimed to reduce the ER to <3 during the medicinal chemistry program, to maximize our chances of achieving efficacious free drug levels in the CNS and other conserved sites.
As 5 showed a suitable profile for hit expansion, we were keen to utilize our cryo-EM platform to generate structural information. Due to challenges with obtaining the T. cruzi proteasome cryo-EM structure, we used the closely related Leishmania tarentolae proteasome as a surrogate (β5 subunit amino acids are 95% similar and 87% identical).
Cryo-Electron Microscopy Studies of 5. The complex of 5 bound to the L. tarentolae 20S proteasome was determined by single-particle cryo-EM to 2.6 Å resolution. Clear features in the electron potential map located at the β4/β5 interface allowed the ligand to be modeled successfully (Supporting Figures 1 and  2).
Compound 5 is bound in a similar orientation to the previously reported 1 17 and 3 25 at the β4/β5 interface close to the catalytic β5Thr100 ( Figure 4). Hydrogen-bond interactions are formed between the central amide with the backbone NH of β5Gly228 and the side-chain phenolic hydroxyl of β5Tyr212. Similar to 1, conformational changes within the β4 subunit occur with respect to the apo L. tarentolae 20S proteasome to accommodate the pyridazinone moiety with a hydrogen bond formed between the pyridazinone carbonyl and the side chain of β5Tyr235. Additional conformational changes are observed in the β4 subunit as the chain of Ile29 rotates, allowing the benzyl moiety to occupy a hydrophobic pocket, forming π−π interactions with side chains of β5Tyr212 and β4Phe24.
To support prospective molecular design, molecules were docked into a homology model of the T. cruzi 20S proteasome developed from the complex of 5 bound to the L. tarentolae 20S proteasome using the homology modeling tool available in Maestro (Schrodinger suites version 2020-04). Docking of 5 to the T. cruzi 20S proteasome homology model replicated the key features of the binding mode defined experimentally for the L. tarentolae 20S proteasome, with the central amide, pyridazinone, and benzyl moieties overlapping as well as the protein residues that define the binding site ( Figure 5). In order to gain a deeper understanding of the ligand−protein interactions driving the molecular recognition process, we carried out a quantum mechanics (MP2-FMO) 35  proteasome. β4 subunit is colored brown, and β5 is colored gray with catalytic β5Thr100 highlighted in red. B: Binding mode of 5 (C atoms gold) compared to 1 (C atoms gray) and bortezomib (C atoms blue). C: Binding mode of 5 (C atoms gold) bound to L. tarentolae 20S proteasome (C atoms gray). Hydrogen bonds are shown in dotted lines. weakened for compound 5. β5Asp116 (−18.18 kcal·mol −1 ) and β5Asp115 (−16.55 kcal·mol −1 ) became the major contributing residues in terms of MP2-FMO PIEDA estimations. Moreover, residues β5Tyr136 (−12.30 kcal·mol −1 ), β4Phe24 (−12.23 kcal·mol −1 ), and β5Tyr113 (−10.89 kcal·mol −1 ) also show improved binding energy contribution, through either hydrogen bonding to the pyridazinone carbonyl (β5Tyr136) or hydrophobic interactions with the benzyl group (β4Phe24 and β5Tyr113), interactions that were not present in compound 1.
Comparing the binding poses of 5 and 1 shows the central amide of 5 overlapping with the urea of 1, the pyridazinone of 5 occupying the same space as the pyrrolidine group of 1 and the phenyl ring of 5 overlapping with the fluorophenyl moiety of 1. The propyl amide group of 5 then occupies some of the space occupied by the bicyclic system of 1. This suggested that transfer of structure−activity relationship (SAR) between the chemotypes might be possible ( Figure 6). This was also supported by the MP2-FMO calculations, which suggested that potency could be increased by improved interaction with β5Thr1 (via extension of the cyclopropyl amide), and by fluorination of the central phenyl ring to increase its dipole and further  strengthen the interaction with β5Asp115 and β5Asp116. The importance of the hydrophobic interactions of the benzyl group also showed that there would be limited value in making diverse changes in this position. Molecular development would therefore focus on fluorination of the central phenyl ring, substitutions to the benzyl moiety extending into an induced fit area of the binding site, and the amide vector extending further into the space occupied by 1.
Structure−Activity Relationship. Equipped with this structural information, starting from hit compound 5, we embarked on a systematic exploration of the molecule to identify a compound suitable for an in vivo proof-of-concept study. The desired compound would have T. cruzi pIC 50 > 6, aqueous solubility > 100 μM, intrinsic clearance <3 mL/min/g (mouse liver microsomes), and ER < 3. We initially explored changes to the central pyridazinone and phenyl groups of 5 (Table 3). Capping the amide nitrogen with a methyl gave 6, with a complete loss of activity confirming that this NH was involved in a key hydrogen-bond interaction with β5Tyr113. Capping the terminal amide nitrogen also led to a significant loss in potency (7). Removing a nitrogen atom from the pyridazinone to give pyridone 8 led to a small drop in potency against the T. cruzi proteasome but a loss of measurable activity in the T. cruzi antiparasitic assay. Both the known SAR from 1 and the output of the MP2-FMO calculations suggested adding a fluorine in the 4-position of the central phenyl ring would result in improved potency. Pleasingly, as predicted, the addition of fluorine in this position to give 9 resulted in >0.5 log unit boost in potency. 9 also showed high aqueous solubility, but the microsomal intrinsic clearance was not suitable for progression to in vivo studies. Moving the fluorine to the 2-position to give 10 was less successful, this compound being 10-fold less potent than 9. Again, this matched the SAR observed in the earlier series. 16 Having identified compound 9 which had improved potency compared to the hit compound 5, we used this as the start-point for the investigation of the benzyl substituent of the pyridazinone with the aim of improving potency and intrinsic clearance while maintaining the high solubility of 5 and 9 ( Table  4). Saturation of the benzyl ring to give the cyclohexyl compound 11 resulted in a loss of potency, likely due to less optimal hydrophobic interactions with β4Phe24 and β5Tyr113. This compound also showed no improvement in microsomal intrinsic clearance. Based on the binding pose, we hypothesized that the addition of a 4-methoxy substituent to 9 could pick up a polar interaction with the side chain of β4Asn22. Pleasingly, this led to 12 with a 0.7 log unit increase in activity against the enzyme which translated to improved activity in the antiparasitic cell-based assay. Microsomal intrinsic clearance was also significantly improved compared to the unsubstituted benzyl compound 5 although solubility was decreased. Moving the methoxy substituent to the 3-position of the phenyl ring to give 13 resulted in a complete loss of potency against the T. cruzi proteasome. Retaining the 4-methoxy substituent and adding a chlorine (14) or a fluorine (15) in the 3-position resulted in a further improvement in potency, with pIC 50 > 8 against the T. cruzi proteasome. 15 was also run in the MDR1-MDCK assay, showing good passive permeability (546 nm s −1 ) and a low ER of 3.0, although it still showed poor solubility (4 μM in the aqueous solubility assay and 11 μM FaSSIF solubility), precluding its progression to in vivo studies. As 14 and 15 showed reasonable intrinsic clearance but poor potency, we investigated alternative 3,4-disubstituted benzyl groups. A similar profile to 14 and 15 was observed for the 4-methoxy-3nitrile compound 16, while 4-chloro-3-fluoro analogue 17 exhibited good potency but was hampered by higher intrinsic clearance compared to the methoxy analogues. Adding a nitrogen into the ring to give 18 had a negative impact on Table 3 a Inhibition of T. cruzi and human proteasome, data from at least three independent replicates, standard deviations ≤0.2. All data are provided in the Supporting Information. b Potency against intracellular T. cruzi amastigotes and host VERO cells, data from at least three independent replicates, standard deviations ≤0.2. All data are provided in the Supporting Information. c Kinetic aqueous solubility measured by a UHPLC system equipped with a UV/visible and single-quadrupole mass spectrometer. d Mouse liver microsomal intrinsic clearance, scaling factor used is 52.5 mg of microsomal protein/g liver. e Chrom Log D pH7.4 = CHI pH7.4 × 0.0857 −2, where CHI is the chromatographic hydrophobicity index. f ND: Not Determined. potency with over a log unit drop compared to the matched pair compound 15. This compound also showed inferior solubility and intrinsic clearance, possibly due to its increased Chrom LogD.
Docking studies using the T. cruzi β4/β5 20S proteasome homology model show very high consistency between the docked poses of 5 and 15, compared to the L. tarentolae cryo-EM complex of 5 ( Figure 7A). Comparison of the MP2-FMO analysis of 5 and 15 in the T. cruzi homology model confirmed the orientation of the benzyl head group, with the 4-methoxy substituent interacting with β4Asn22 (indicated by a −1.16 kcal· mol −1 improvement (ΔΔG) in energy contribution) as expected, and also improving the interaction with β4Phe24 (ΔΔG = −4.38 kcal·mol −1 ). Also as expected, the fluorine oriented toward β4Val121 (ΔΔG = −3.45 kcal·mol −1 ), while also giving a more significant energetic contribution with β5Asp115 (ΔΔG = −7.99 kcal·mol −1 ) within a total calculated binding energy improvement of −18.75 kcal·mol −1 going from 5 to 15. Interestingly, the contribution from β4Phe24 and β5Asp115 is mainly through electrostatic interactions, rather than any other forms of energetics ( Figure 7E). This corresponds well with electrostatic potential (ESP) analysis of both the protein and the compounds ( Figure 7B−D) where the methoxy and fluoro substituents of 15 induce an uneven electrostatic distribution about the benzyl group, with partial positivity ( Figure 7D) exposed to the side chains of β4Phe24 and β5Asp115, which both maintain partial negativity ( Figure  7B). This electrostatic complementarity is likely facilitating the activity boost, alongside the dispersive component of β4Phe24.
Comparing the binding modes of 5 and 1 suggested there was an opportunity to expand the amide vector, with the MP2-FMO calculations showing that there might be the opportunity to pick up an interaction with β5Thr1. With the optimized benzyl substituents of 12 (4-methoxy) and 15 (3-fluoro-4-methoxy), Table 4 a Inhibition of T. cruzi and human proteasome, data from at least three independent replicates, standard deviations ≤0.2. All data are provided in the Supporting Information. b Potency against intracellular T. cruzi amastigotes and host VERO cells, data from at least three independent replicates, standard deviations ≤0.2. All data are provided in the Supporting Information. c Kinetic aqueous solubility measured by a UHPLC system equipped with a UV/visible and single-quadrupole mass spectrometer. d Mouse liver microsomal intrinsic clearance, scaling factor used is 52.5 mg of microsomal protein/g liver. e Chrom Log D pH7.4 = CHI pH7.4 × 0.0857 − 2, where CHI is the chromatographic hydrophobicity index. f Number in parentheses refers to FaSSIF solubility (fasted state simulated intestinal fluid solubility, μM). g ND: Not Determined. h Inactive against Vero cells in seven replicates, potency shown is from one active replicate.
we investigated alternative substitutions on the amide, aiming to retain (or improve) potency while improving solubility and intrinsic clearance. To explore this vector, we synthesized both 4-methoxy and 3-fluoro-4-methoxybenzyl analogues, as the latter generally led to more potent, but less soluble, compounds, and it was important to identify the molecules with the best overall profiles ( Table 5). The 3-methoxy cyclobutene analogues 19 and 20 showed good potency and intrinsic clearance but were hampered by poor solubility. Morpholine analogues 21 and 22 showed much improved solubility profiles; however, intrinsic clearance was higher than desired and 22 was found to be a Pgp substrate when assessed in the MDR1-MDCK assay (ER 13.3). The 3-morpholino-cyclobutyl analogues 23 and 24 exhibited high potency. While cis isomer 24 suffered from low solubility, trans isomer 23 showed a good balance of properties but was found to be a Pgp substrate with ER of 22.2. Amino alcohol compound 25 also showed good potency against the target and in the whole cell assay as well as exhibiting good solubility and low intrinsic clearance. Unfortunately, this compound also had low passive permeability in the MDR1- MDCK assay (133 nm s −1 ) with a high ER of 21.2. As the extended analogues were not improving potency and did not deliver compounds with suitably balanced properties for progression to in vivo studies, we investigated the effect of truncating the amide moiety. Methyl amide 26 maintained similar potency to the more elaborated analogues and had very good overall properties with high passive permeability, an ER of 1.8, and FaSSIF solubility measured at 339 μM, making this compound of particular interest. As expected, fluorination of the phenyl ring of 26 to give compound 27 increased potency in both the biochemical and cellular assays, as well as improving intrinsic clearance. Unfortunately, 27 was significantly less soluble than 26, so it was not progressed further.
For the synthesized compounds, a plot of proteasome inhibition against antiparasitic activity showed a clear correlation (Figure 8), with a relatively consistent 10-fold drop-off in activity against the parasite, suggesting that the antiparasitic activity was being driven by proteasome inhibition. Also, a plot of Chrom Log D vs T. cruzi proteasome inhibition ( Figure 9) showed that it was possible to increase potency from the initial hit, while reducing lipophilicity. This focus on compounds that increased LLE by reducing lipophilicity was critical to identifying compounds that coupled good levels of potency with desirable solubility and metabolic stability.
In Vivo Studies on 15 and 26. From the chemistry program, compounds with suitable profiles to progress to in vivo studies were identified. Initially, we investigated 15 due to its high intracellular potency, low intrinsic clearance in mouse liver microsomes, and reasonable permeability. The ER of 6, coupled with high passive permeability (Papp(A−B) > 300 nm/s) was an improvement on compound 4, although aqueous solubility in both buffer and FaSSIF media was low. When dosed orally at 50 mg/kg in mice, 15 gave total blood levels above EC 99 for around 14 h (Figure 10), although accounting for the blood−brain ratio of 0.42 it was predicted that total brain concentration would only reach the EC 99 for approximately 12 h, with the unbound brain concentration never reaching EC 99 . Table 5 a Inhibition of T. cruzi and human proteasome, data from at least three independent replicates, standard deviations ≤0.2. All data are provided in the Supporting Information. b Potency against intracellular T. cruzi amastigotes and host VERO cells, data from at least three independent replicates, standard deviations ≤0.2. All data are provided in the Supporting Information. c Kinetic aqueous solubility measured by a UHPLC system equipped with a UV/visible and single-quadrupole mass spectrometer. d Mouse liver microsomal intrinsic clearance, scaling factor used is 52.5 mg of microsomal protein/g liver. e Chrom Log D pH7.4 = CHI pH7.4 × 0.0857 −2, where CHI is the chromatographic hydrophobicity index. f Permeability in MDR1-transfected MDCK cells in the presence of a Pgp inhibitor (GF120918)/efflux ratio. g Number in parentheses refers to FaSSIF solubility (fasted state simulated intestinal fluid solubility, μM). h ND Not Determined. i Inactive against Vero cells in 9 out of 10 replicates, data for active replicate shown.
We therefore proceeded to examine 26, which had similar potency to 15. As seen for 4, 26 performed well in our in vitro washout assay, with relapses at 35 and 42 days post washout compared to 31 and 35 days for BNZ (two replicates each, 16day treatment at 25-fold EC 50 ). Although intrinsic clearance in mouse liver microsomes was poorer, 26 did show much improved solubility and permeability, a lower ER of 1.8, and a brain−blood ratio of 0.70. 26 was therefore progressed into a pharmacokinetic (PK) study, dosed at 50 mg/kg, and was seen to have total blood exposure exceeding the EC 99 for >8 h, with unbound exposure above EC 99 for approximately 7 h ( Figure  11). Based on the brain−blood ratio, and the measured brain tissue binding (13.3%), brain total concentration was likely to exceed EC 99 for around 7 h, with brain unbound concentration above EC 99 for around 3 h.
Based on this data, we considered 26 to be a suitable compound to give proof of concept in the chronic mouse model of Chagas disease. We elected to dose at 50 mg/kg b.i.d for 20   However, by day 20, this had dropped to less than 2 h after the first dose ( Figure 12). 26 showed no effect on mouse PXR, even at 30 μM, so autoinduction due to PXR-mediated upregulation of CYP 450 enzymes was unlikely to be the explanation.

Journal of Medicinal Chemistry
The ability of 4 and 26 to outperform BNZ in our in vitro washout assay, alongside the results reported for 2, suggests that proteasome inhibition is a valid strategy for targeting T. cruzi. Unfortunately, we were unable to translate this into efficacy in a chronic mouse model of Chagas disease for 26. The most likely explanation is that blood-free levels were not maintained above the EC 99 for the total dosing duration. Furthermore, the brain exposure data for 26 indicated that it also may not reach all key tissues with sufficient levels of free drug. During washout assays exposure is kept high (>EC 99 ) continuously for the duration of treatment, while in the in vivo studies, free compound exposure  for 26 dropped below EC 99 after ∼5 h post-dosing on day 1 and even sooner on day 20, potentially allowing the parasites to recover each day. This issue could potentially have been addressed using higher doses of 26 to increase exposure, but a dose escalation study was run which showed that 26 was not tolerated at higher doses (data not shown). Unfortunately, this meant that our hypothesis of utilizing CNS exposure to predict potential efficacy was not fully tested, and whole-body imaging experiments are currently ongoing to better understand the in vivo distributional characteristics of key compounds within the series compared to parasite localization. This will be reported in a subsequent publication.
Synthetic Chemistry Section. To facilitate the synthesis of the described compounds, we first synthesized a set of key building blocks. Initially, alkylated pyridazinone 3-carboxylic acids were synthesized from methyl 6-oxo-1,6-dihydropyridazine-3-carboxylate, starting by alkylation of the pyridazinone nitrogen with a suitable alkyl halide to give 28 (a−i) with subsequent ester hydrolysis yielding the acids 29 (a−i) (Scheme 1). Alongside these, anilines 31, 33, 34, and 35 were synthesized. Initially, BOC-protected 3-aminobenzoic acid was coupled with cyclopropylamine using T3P to give 30, with subsequent BOC deprotection yielding aniline 31. Alternatively, 2-fluoro-5nitrobenzoic acid was coupled to cyclopropylamine using T3P to yield 32, which was reduced with iron/ammonium chloride to yield aniline 33. Finally, 5-amino-2-fluorobenzoic acid was coupled directly with trans-3-methoxycyclobutan-1-amine using EDC/HOBt to give 34, or coupled with methylamine using T3P/N,N-diisopropylethylamine (DIPEA) to give 35. These intermediates could be utilized for the synthesis of the target compounds following two different routes.
Following a linear route, the benzoic acids 29 (a, c, f) could be coupled with suitable anilines via T3P or HATU to give amides 36 (a−e) (Scheme 2). To generate the methylated amide, 36a was treated with iodomethane and potassium carbonate to give

■ CONCLUSIONS
Based on our in vitro data package, alongside previously reported work, 24 inhibition of the proteasome appears to be an attractive mechanism of action for new drugs for the treatment of Chagas disease. Our previously identified series of Leishmania proteasome inhibitors failed to provide compounds with activity in a chronic mouse model of Chagas disease, which we hypothesized was due to a disconnection between the distribution of compound vs parasite in vivo. We therefore undertook a hit-discovery program which led to the identification of 5 as an attractive hit compound. A cryo-EM structure of 5 bound to the L. tarentolae proteasome showed that, like the previous series, it bound at the β4/β5 interface close to the catalytic β5Thr100, although it also occupied an additional hydrophobic pocket. Utilizing the structural information, hit expansion of 5 led to the identification of two compounds, 15 and 26, with suitable profiles to progress to in vivo studies. 26 proved to have a much more attractive in vivo profile compared to 15 and was therefore progressed into a mouse efficacy model of chronic Chagas disease, but disappointingly was not curative. The reasons for this were likely related to blood exposure not being maintained above EC 99 for the whole dosing interval, and limited distribution to all of the sites where the parasite resides. Although this series did not deliver efficacious compounds, proteasome inhibition remains an attractive strategy for targeting Chagas disease and other kinetoplastid-related diseases. To help understand if compounds are distributed to locations where the parasites are residing within the mouse, whole-body imaging experiments are ongoing, which will map compound distribution to where the T. cruzi parasites reside. If the compound is not distributed to all of the sites where the parasites reside, this could explain the lack of efficacy of 26. These studies will help in the design of the next generation of T. cruzi proteasome inhibitors and will be reported in a subsequent publication.
■ EXPERIMENTAL SECTION Chemistry. Chemicals and solvents were purchased from the Aldrich Chemical Company, Fluka, ABCR, VWR, Acros, Fluorochem, and Alfa Aesar and were used as received. Air-and moisture-sensitive reactions were carried out under an inert atmosphere of nitrogen in oven-dried glassware. Flash column chromatography was performed using pre-packed silica gel cartridges (230−400 mesh, 40−63 μm, from Redisep) using a Teledyne ISCO Combiflash Companion, or Combiflash Retrieve. 1 H NMR and 13 C NMR spectra were recorded on a Bruker Avance DPX 500 spectrometer ( 1 H at 500.1 MHz, 13 C at 125.8 MHz). Chemical shifts (δ) are expressed in ppm recorded using the residual solvent as the internal reference in all cases. Signal splitting patterns are described as singlet (s), doublet (d), triplet (t), quartet (q), multiplet (m), broad (br), or a combination thereof. Coupling constants (J) are quoted to the nearest 0.1 Hz. High-resolution electrospray measurements were performed on a Bruker Daltonics MicrOTOF mass spectrometer. Low-resolution electrospray (ES) mass spectra were recorded on an Advion Compact mass spectrometer (CMS: model ExpressIon CMS) connected to Dionex Ultimate 3000 UPLC system with diode array detector, or an Acquity UPLC (MS: Waters SQD; ELSD: Waters 2424; Waters PDA; Waters Binary solvent manager; Waters sample manager). HPLC chromatographic separations were conducted using a Waters XBridge C 18 column (2.1 mm × 50 mm, 3.5 μm particle size) or Waters XSelect column (2.1 mm × 30 mm, 2.5 μm particle size), eluting with a gradient of 5−95% acetonitrile/water +0.1% ammonia or +0.1% formic acid, or a Waters Acquity BEH C 18 column (3 mm × 50 mm, 1.7 μm particle size) eluting with a gradient of 5−95% acetonitrile/water +0.1% formic acid. All intermediates had a measured purity ≥90% and all assay compounds had a measured purity of ≥95% as determined using analytical LC-MS (TIC and UV). The synthesis of all intermediates, and spectral data for final compounds, is included in the Supporting Information.