Development of (6R)-2-Nitro-6-[4-(trifluoromethoxy)phenoxy]-6,7-dihydro-5H-imidazo[2,1-b][1,3]oxazine (DNDI-8219): A New Lead for Visceral Leishmaniasis

Discovery of the potent antileishmanial effects of antitubercular 6-nitro-2,3-dihydroimidazo[2,1-b][1,3]oxazoles and 7-substituted 2-nitro-5,6-dihydroimidazo[2,1-b][1,3]oxazines stimulated the examination of further scaffolds (e.g., 2-nitro-5,6,7,8-tetrahydroimidazo[2,1-b][1,3]oxazepines), but the results for these seemed less attractive. Following the screening of a 900-compound pretomanid analogue library, several hits with more suitable potency, solubility, and microsomal stability were identified, and the superior efficacy of newly synthesized 6R enantiomers with phenylpyridine-based side chains was established through head-to-head assessments in a Leishmania donovani mouse model. Two such leads (R-84 and R-89) displayed promising activity in the more stringent Leishmania infantum hamster model but were unexpectedly found to be potent inhibitors of hERG. An extensive structure–activity relationship investigation pinpointed two compounds (R-6 and pyridine R-136) with better solubility and pharmacokinetic properties that also provided excellent oral efficacy in the same hamster model (>97% parasite clearance at 25 mg/kg, twice daily) and exhibited minimal hERG inhibition. Additional profiling earmarked R-6 as the favored backup development candidate.


■ INTRODUCTION
Visceral leishmaniasis (VL) is a particularly lethal sandfly-borne parasitic disease that is prevalent in more than 60 countries, where it mostly affects underprivileged people in remote rural areas who have limited access to diagnosis and treatment. 1−3 Major outbreaks of VL in East Africa have been attributed to waves of forced migration during periods of conflict, and such epidemics are exacerbated by weak healthcare systems, malnutrition, and HIV/AIDS coinfection. 4,5 Moreover, in this region, the first-line drug combination of paromomycin and sodium stibogluconate was found to be unsuitable for VL patients who were >50 years of age or those with HIV, and no other therapies have shown adequate efficacy. 6,7 Failure of the most recently evaluated new agent, fexinidazole, in a phase II clinical trial for VL in Sudan 8 has now left the clinical pipeline empty, underlining the compelling need to develop more satisfactory medications. 9 The target product profile (TPP) of an optimized new chemical entity for the treatment of VL requires (i) effectiveness against all causative species, in all endemic areas, in both immunocompetent and immunosuppressed individuals, with a clinical efficacy of >95%; (ii) activity against resistant strains; (iii) no adverse safety events requiring monitoring and no contraindications; (iv) no drug−drug interactions (suitable for combination therapy); (v) oral administration once per day for a maximum of 10 days (or intramuscular dosing three times over 10 days); (vi) stability in relevant climates (3 years); and (vii) affordable cost (<$80, ideally <$10 per course). 10 However, new drug discovery for VL faces formidable challenges, such as inadequate investment, a lack of validated targets, poor translation of in vitro activity into in vivo models, and meager hit rates (<0.1%) for phenotypic screening of compound libraries. 11−13 The latter may be due in part to the concealed location of parasites in acidic parasitophorous vacuoles within macrophages. 14 Furthermore, the unique glycolipid-rich cell surface of the amastigotes presents an additional barrier to chemotherapy. 15 Another issue is that many cellularly active hits may never meet TPP and progression criteria, even after valiant optimization attempts. 13,16 Nevertheless, drug development efforts spearheaded by the Drugs for Neglected Diseases initiative (DNDi) have now shown encouraging progress in several novel classes, including oxaboroles and aminopyrazoles. 12,17,18 Novartis has also disclosed a triazolopyrimidine preclinical lead with utility in vivo against both leishmanial and trypanosomal infections. 19 The 2-nitroimidazooxazines are best known for their potent effects against Mycobacterium tuberculosis (M. tb), the causative agent of tuberculosis (TB). 20 The first drug candidate from this class, pretomanid [PA-824, S-1 (Figure 1)], has shown excellent safety and bactericidal efficacy in phase II clinical trials for TB, 21 leading to its ongoing combination assessment, 22 while our collaborative work with the TB Alliance on second-generation analogues of S-1 culminated in the advancement of TBA-354 (S-2) into phase I studies. 23 We have recently described the investigation of a novel 7-substituted 2nitroimidazooxazine class, which in addition to possessing considerable potential against TB has also demonstrated exciting activity against both VL and Chagas disease, resulting in the selection of preclinical VL lead 3. 24 This followed an indepth analysis of the structurally related 6-nitroimidazooxazole class, 25 where phenotypic screening of some of our initial examples by DNDi had enabled the discovery of previous development nominee 4 (DNDI-VL-2098). 26 The latter was found 27 to be activated by a novel leishmanial nitroreductase (NTR2). In comparison to 4, candidate 3 exhibited an improved safety profile and had similarly notable efficacy in two animal models of VL. 24 Furthermore, while the new TB drug delamanid (5) has also been suggested as a possible VL therapy, 28 it is noteworthy that 3 was substantially more effective than this agent in the highly stringent chronic infection hamster model. 24,25 As part of our VL lead optimization program with DNDi, it was considered important to develop a few efficacious backup compounds having good physicochemical/pharmacological profiles and better safety, to mitigate development risks. Given the encouraging results with nitroimidazooxazoles and 7-substituted 2-nitroimidazooxazines, we first evaluated various other pretomanid-related scaffolds for VL, including those with a reversed linker at C-6 29,30 and novel nitroimidazooxazepines. We then assessed our larger collection of pretomanid analogues via the medium-throughput screening of ∼900 compounds at the Institut Pasteur Korea (IPK). Finally, a more systematic synthetic approach was employed to redevelop the 6substituted 2-nitroimidazooxazine class for VL, taking into consideration both enantiomer forms. We now report the findings from these wide-ranging structure−activity relationship (SAR) studies, including the detailed in vitro/in vivo profiling of selected new leads, which resulted in our identification of the title compound as a very promising VL backup candidate.
■ CHEMISTRY Scheme 1 outlines the synthetic methods used to prepare eight novel racemic analogues of S-1 featuring changes to the original nitroimidazooxazine core (14, 22, 23, 32, 33, 39, 40, and 47). A common strategy (based on the original, well-validated route to S-1 and simple derivatives) 31,32 proved to be effective for the first five of these (Scheme 1A−C), involving the initial reaction of functionalized epoxides (9, 33 17, and 27 34 ) with 2,4dinitroimidazole (8), followed by THP protection of the derived alcohols (10, 18, and 28). In the shorter chain cases (11 and 19), subsequent cleavage of the TBS ether (TBAF) enabled in situ annulation, whereas in the latter instance (29), oxazepine ring formation required additional treatment with a strong base (NaH). Removal of the THP group with methanesulfonic acid and standard alkylation chemistry on alcohols 13, 21, and 31 then gave the aforementioned targets.
For the isomeric oxazepines (Scheme 1D), THP protection of alcohol 34 24 and desilylation (TBAF) similarly furnished the noncyclized alcohol 36, which was ring-closed (NaH and DMF) and THP-deprotected (HCl) to produce alcohol 38. However, both final step alkylations of 38 unexpectedly cogenerated significant quantities (20−36%) of the isomeric 7-substituted oxazine derivative (41 or 42), together with the desired oxazepine ether (39 or 40). Although this is unconfirmed, it is postulated that this rearrangement may involve an intramolecular S N Ar reaction, with attack by the oxazepine alkoxide anion at the imidazole ring junction carbon 9a and subsequent alkylation of the released 2-nitroimidazooxazine 7-alkoxide (but it should be emphasized that no similar rearrangements were detected during the derivatization of 13, 21, 31, or the alcohol precursor to S-1, and that S-1 itself has shown excellent safety and a lengthy 16−18 h half-life in clinical trials for TB, 35 suggesting that 2-nitroimidazooxazine-based VL leads would be unlikely to demonstrate an excessive reactivity toward biological nucleophiles; this is further supported by an observed tolerance of the latter ring system toward several basic nucleophiles in the chemistry reported below). The remaining
Several reference benzyl ethers (R-1, 38 R-7, 32 R-118, and R-119) were sourced through direct alkylations of 6-R alcohol 56 32 (Scheme 4A). Three extended ether targets (R-122, R-123, and R-124) were also formed by Sonogashira reactions on the propargyl ether R-121, derived from the coupling of 56 with bromide 120. 41 Next, the orthogonally diprotected triol 125 42 was employed in complementary syntheses of the novel R and S enantiomers of racemic ether 6 42 (Scheme 4B,C). Following a Mitsunobu reaction of 125 with 4-(trifluoromethoxy)phenol, selective removal of the PMB group (DDQ), iodination of the resulting alcohol 127, and reaction with 2-bromo-4-nitroimidazole (129) gave silyl ether 130. Treatment of the latter with TBAF and sodium hydrideinduced ring closure then produced S-6. Conversely, successive cleavage of the TIPS group from 126, iodination, and then reaction with 129 provided PMB ether 134. Oxidative debenzylation of 134 with DDQ in the absence of water unexpectedly led to partial acetalization of the alcohol product with 4-methoxybenzaldehyde, but this mixture was cleanly converted back to 135 via acid hydrolysis (TsOH/MeOH). Base-assisted annulation of alcohol 135 (NaH) then furnished the second enantiomer, R-6. In subsequent work, 2-pyridinyl ether analogues of both R-6 (R-136, R-137, and R-138) and triazole 47 (139) were accessed from alcohols 56 32  Further linker diversity was accessed through the transformation of 6-S alcohol 65 32 into the novel 6-R amine hydrochloride, 146 (Scheme 5A). Following tosylation of 65 and azide displacement, reduction of 6-R azide 145 with propane-1,3-dithiol gave the required amine, which was converted to its hydrochloride salt for improved stability. From this intermediate, reductive alkylation with benzaldehydes (using NaBH 3 CN), acylation with benzoyl chlorides, or treatment with phenyl isocyanates in the presence of catalytic dibutyltin diacetate yielded the expected benzylamine, benzamide, or phenyl urea derivatives (Scheme 5A,B). Then, to conclude this study, a variety of shorter O-linked heterobiaryl side chains were constructed using Suzuki couplings on haloheteroaryl ether precursors [156,159,167,170, and 173 (Scheme 5C−E)]. The latter were obtained directly from alcohol 56, via S N Ar reactions on fluoropyridines or chloropyrimidines, or, in the case of 167, from the diprotected triol 125 42 and 6-bromopyridin-3-ol, using the same methodology as described above for R-6.

■ RESULTS AND DISCUSSION
To establish the SARs against kinetoplastid diseases, 76 new (and several known) pretomanid analogues derived from successive projects with TB Alliance and DNDi were retrospectively tested in replicate assays conducted at the University of Antwerp [LMPH (Tables 1−4)]. These assays measured activity versus the intracellular amastigote forms of both Leishmania infantum (L. inf) and Trypanosoma cruzi (T. cruzi) and against the bloodstream form of Trypanosoma brucei (T. brucei); cytotoxicity toward human lung fibroblasts (MRC-5 cells, the host for T. cruzi) was also assessed. 44 Much of our VL lead optimization work with DNDi was earlier guided by the findings from single IC 50 determinations against Leishmania donovani (L. don) in a mouse macrophage-based luciferase assay 26 performed at the Central Drug Research Institute (CDRI, Pradesh, India), and by follow-up evaluations of in vitro microsomal stability and efficacy in the mouse VL model (Figure 2A). The best leads were then advanced to further appraisal in the more stringent hamster VL model. Overall, while excellent in vivo efficacy was a key goal for secondgeneration VL drug candidates, we also aspired (a) to minimize compound lipophilicity (estimated using ACD LogP/LogD software, version 14.04, Advanced Chemistry Development Inc., Toronto, ON) to lessen toxicity risks, (b) to increase aqueous solubility (as judged by kinetic data on dry powder forms of active leads) for optimal oral bioavailability, and (c) to reduce hERG inhibition potential (cf. 4) 24 to improve safety.
Scaffold Modification: Initial Hits. As part of our earlier TB studies, we had examined some fundamental changes to the nitroimidazooxazine "warhead", including replacement of the nitroimidazole ring by nitropyrazole or nitrotriazole [e.g., 48-50 (Table 1)] and exchange of the 8-oxygen for sulfur or nitrogen. 40 We had also explored reversal of the C-6 linker (e.g., 24−26) 29,30 and transposition of the side chain to position 7 (e.g., 41 and 42). 24 In a further extension to this work (seeking improved metabolic stability and new active

Journal of Medicinal Chemistry
Article scaffolds), the novel methylated derivatives 14, 22, and 23 were investigated, together with nitroimidazooxazepines 32, 33, 39, and 40. Unfortunately, except for the 7,7-dimethyl derivative 22 [which showed antitubercular potency comparable to that of 1 and excellent stability toward human liver microsomes, HLM, 92% after 1 h (Table 5)], these compounds proved to be unattractive for TB. Nevertheless, in preliminary antiparasitic screening at the Swiss Tropical Institute, 23, 25, and 26 demonstrated encouraging utility against L. don in a mouse macrophage assay (IC 50 s of 1.2−1.5 μM), 29 and triazole 49 exhibited striking activity versus Chagas disease (T. cruzi IC 50 of 0.084 μM). For greater clarity, we will focus the discussion first on the intended main application (VL) and discuss the other parasite data in a closing section. Follow-up testing of a larger set of compounds at CDRI 26 identified 1, 6, 24, and 40 as being superior for VL [L. don IC 50 s of 0.26−0.46 μM (Table 1)], although such hits were still an order of magnitude less potent than 4 and the 7-substituted oxazine 41 (IC 50 s of 0.03 μM). 24,25 These results, together with evidence of the reduced solubility and more rapid metabolism of analogues with 4benzyloxybenzyl side chains, 36,42 prompted the further appraisal of 24 in an L. don infection VL mouse model. Disappointingly, 24 displayed weak activity [31% inhibition at 25 mg/kg, dosing po daily for 5 days (Table 5)], despite its reasonable mouse PK profile [50% oral bioavailability, moderate exposure, and 2 h half-life (Table 6)]. This outcome implied the need to significantly boost in vitro potencies in this class. However, two highly effective phenylpyridine analogues of 24, 116 30 Table 2)], also failed to deliver useful activity in this in vivo assay under the same dosing regimen (23−49% inhibition). Analysis of their mouse PK data identified low oral bioavailability (11−15%) as a contributing factor here because greater oral exposure led to better efficacy [116 (Table 6 and Figure S1)]. A related concern for both compounds was poor aqueous solubility [0.13−0.27 μg/mL at pH 7 (Table 5)], while retrospective testing against L. inf later revealed suboptimal potency (IC 50 s of ∼6 μM). Taken together, these findings reinforced the importance of improving both potency and in vivo PK properties, to achieve suitable efficacy in VL models. To this end, we returned to our sizable pretomanid analogue library, where we had already amassed key solubility and DMPK information from extensive earlier studies with the TB Alliance.
Library Screening and Hit to Lead Assessments for VL. To assist the identification of more active leads from the 6substituted nitroimidazooxazine class, an 898-member library was screened against L. don amastigotes in a seven-point 3-fold dilution macrophage assay at the Institut Pasteur Korea (mid-2010). 45 In total, 248 compounds (28%) showed >50% inhibition at 10 μg/mL, although only 89 (36%) of these displayed >50% inhibition at 3.3 μg/mL, and known actives from the nitroimidazooxazole and 7-substituted oxazine classes were dispersed across both groupings. By eliminating examples from previously inspected classes (including the "reversed C-6 linker" series above), we obtained a starting set of 169 hits. This set was further refined by excluding compounds with a higher propensity for metabolic and/or solubility issues, based on established trends, to give 42 hits, which were retested at CDRI.

Scheme 3 a
The most relevant results are summarized in Figure 3, with almost all of the remaining hits manifesting weaker potencies (L. don IC 50 s of 0.5−11 μM). We also screened S-1 and S-2, but both had only modest activities (L. don IC 50 s of 3.9 and 2.6 μM, respectively). Nevertheless, in view of the 10-fold higher potency of racemic 1 [IC 50 of 0.39 μM (Table 1)], this result for S-1 was highly significant as it implied that 6R enantiomers (which have little activity against M. tb 32,37 ) may be the more active chiral form for VL. Therefore, we synthesized a trial set of 10 compounds (R-1, R-2, R-7, R-58, R-60, R-62, R-77, R-81, R-84, and R-94) for assessment. In 9 of 10 cases, these 6R enantiomers exhibited 1.  Tables S2 and S3)]; hence, the 6R counterparts of selected hits in Figure 3 were also targeted (in line with the approach in Figure 2A).
In an effort to further prioritize the library screening hits for in vivo evaluation, five compounds of high lipophilicity (CLogP > 4 for S-53, 46 S-54, 30 S-55, 47 S-58, 47 and S-64 39 ) were omitted from further study and the remaining 12 were assessed for aqueous solubility and microsomal stability (Table 5 and cited references for Figure 3). Both the amide S-151 48 and urea S-155 provided encouraging solubility data (132 and 22 μg/mL, respectively), but the urea unexpectedly showed poor stability toward mouse liver microsomes (MLM, 43% parent after 30 min). Conversely, both the lipophilic arylthiazole S-51 49 (CLogP ∼ 4.0) and the arylpyrimidine S-171 42 were considered of borderline interest because of their modest solubility values (0.9−1.6 μg/mL). While some phenylpyridine hits (e.g., S-77, S-89, S-91, S-92, and S-99) 39 were not substantially more soluble than this at pH 7 (1.4−4.0 μg/mL), these compounds have demonstrated greatly superior results at pH 1 (211−1780 μg/mL). It has been recognized that the low pH of gastric fluid (typically ∼1−2) can enhance the dissolution and oral absorption of such weak bases. 50 Furthermore, close analogue S-2 was advanced to clinical studies for TB partly on the basis of its superior in vivo PK properties in comparison to those of delamanid (5), which are absorption-limited. 23 Concordant with this, the most potent phenylpyridine hits in Figure 3 (S-77, S-81, S-89, S-91, and S-92) also displayed broadly acceptable HLM and MLM stabilities (>70% remaining after 30 min), suggesting their suitability for in vivo studies.

SAR of 6-Substituted 2-Nitroimidazooxazines for VL.
Following the discovery that many 6R enantiomers had superior in vitro and in vivo activities against VL, a more extensive lead optimization study was initiated to develop additional backup candidates to 4 possessing an advantageous solubility, PK−PD, and safety profile. In light of the high potency of ortho-linked biphenyl hit S-58 (L. don IC 50 0.18 μM), we first sought to establish the optimal linking position for biaryl side chains. Comparison of R-58, R-60, and R-62 in both L. don and L. inf assays ( Table 2) unexpectedly indicated that ortho linkage was most preferred and that para linkage was least preferred. Therefore, the novel ortho-linked phenylpyridines S-69, R-69, S-74, and R-74 were studied. Here, R-69 and R-74 were equally best, although 1.8-fold less effective than R-58 (L. inf IC 50 s of 2.0 vs 1.1 μM). Interestingly, these two phenylpyridine isomers showed major differences in both solubility and microsomal stability, with the more soluble R-74 (78 vs 0.51 μg/mL) being metabolized extremely rapidly in all three microsome species [0.1−8% remaining after 1 h (Table  5)], whereas R-69 was more stable than the para-linked analogue R-84 described above (44% vs 36% after 1 h in MLM).
In the meta-linked phenylpyridine series, two new compounds (R-79 and R-80) having terminal ring substituents
As suggested by the screening data ( Figure 3), replacement of the proximal phenyl ring in R-62 with 3-pyridine (R-2) was less favorable [L. inf IC 50 of 4.1 μM vs 0.71 μM for R-84 (Table  2)]. Nevertheless, the 6R enantiomers of two hits (R-99, 2-F, 4-OCF 3 ; R-101, 2-Cl, 4-OCF 3 ) and the novel 2,4-difluoro analogue R-106 all displayed good potencies (L. inf IC 50 s of 1.1, 0.61, and 0.85 μM, respectively) and microsomal stabilities at least comparable to those of their 2-pyridine counterparts, although R-101 was cytotoxic (MRC-5 IC 50 of 17 μM). By extension, we examined three less lipophilic diaza proximal rings (R-109, R-112, and R-115) that had proven to be very effective in our TB studies, 39 but these turned out to be of less interest (L. inf IC 50 s of 1.4−2.8 μM). In summary, several new phenylpyridines provided profiles that were attractive for in vivo evaluation, but we had yet to investigate other linker groups. Therefore, we next turned our attention to simpler monoaryl side chains to explore these changes. For this part of the study, we restricted our focus to linkers that had shown particular promise either in the initial screening (e.g., 6-O, 6-NHCO, and 6-NHCONH) or in our earlier TB work.
Commencing with the enantiomer of pretomanid [R-1, L. inf IC 50 of 4.7 μM (Table 3)], we found variation of the trifluoromethoxy position identified that ortho substitution was      (Table 5)]. Therefore, we similarly investigated trifluoromethylpyridine analogues of R-6 (R-136, R-137, and R-138) and while there was an ∼2-fold loss of activity, R-136 demonstrated a 9-fold  Another option to improve solubility was to replace the ether linkage at C-6 with nitrogen-based linkers. 48 The 6-amino analogue of R-1 (R-147) had a 5-fold better solubility value at pH 7 (84 μg/mL) and was >2000 times more soluble at pH 1. Nevertheless, this compound was less stable toward microsomes (e.g., 61% vs 86% in MLM) and showed 2.6-fold lower activity (L. inf IC 50 of 12 vs 4.7 μM), which was not sufficiently improved by varying the ring substituent position (R-148 and R-149). Alternatively, with a carboxamide or urea linker, the enantiomer preference was reversed, with the original 6S forms (S-150, S-151, S-152, S-154, and S-155) being clearly superior but not particularly potent (L. inf IC 50 s of ∼6−10 μM). Hence, the only compounds with useful antileishmanial activity were ether-linked at C-6, although it was apparent that the original OCH 2 linkage was not optimal and that removal of the benzylic methylene may have metabolic stability and potency advantages. To investigate this further, a small set of O-linked phenylpyridine and phenylpyrimidine derivatives was evaluated (Table 4). In the phenylpyridine series (R-157 to R-169), activity against L. inf was similar to or better than that for R-6, with the 3-pyridine isomers preferred (R-168 and R-169, IC 50 s of 0.13−0.20 μM). These latter compounds exhibited microsomal stabilities comparable to those of the parent linker series above, but their solubility values were inferior (0.36−1.2 μg/ mL at pH 7). Finally, contrary to the screening data for S-171, a proximal pyrimidine ring was tolerated only when it was paralinked (R-174 and R-175, L. inf IC 50 s of 0.65−0.83 μM). Generally, potency against L. inf was more discriminating and tended to better correlate with in vivo outcomes. 24,25 Throughout the course of these studies, 10 more candidates (R-1, R-6, R-69, R-96, R-99, R-106, R-136, R-147, R-168, and R-169) were screened for activity in the mouse VL model at 50 mg/kg [dosing po daily for 5 d (Table 5)]. The ortho-linked phenylpyridine R-69 was disappointing (20% inhibition), but the new para-linked analogues R-96, R-99, and R-106 displayed high efficacies (91−97%). More heartening still was the fact that shorter chain phenylpyridines R-168 and R-169 gave essentially complete clearance of the parasite infection (99.9%), as did their monoaryl counterparts, R-6 and R-136. However, both R-1 and its amino-linked equivalent R-147 were unsatisfactory (50 and 12%, respectively), consistent with their weaker in vitro potencies. During the concluding stages of this project, Patterson et al. 38 reported that R-1 was a potential oral treatment for VL on the basis of its in vivo activity in a comparable mouse model at a much larger dose of 100 mg/kg twice daily, but it is clear from these results and other studies 24 that R-1 may not be the optimal development candidate.
Additional Assessments To Determine the Best VL Lead. Dose−response experiments on 7 of the 10 best compounds (R-6, R-84, R-89, R-91, R-92, R-96, and R-99) in the mouse VL model yielded ED 50 values of 7.5, 12, 14, 28, 20, 28, and 19 mg/kg, respectively (Table 5 and Figure 4). The more recent O-linked leads, R-136, R-168, and R-169, were also evaluated at a smaller dose of 6.25 mg/kg and provided parasite burden reductions of 30, 99.7, and 72%, respectively. The most efficacious of these (R-168) produced 84% inhibition at 3.13 mg/kg, representing activity at a level similar to that of

Journal of Medicinal Chemistry
Article the nitroimidazooxazole 4, 25,26 albeit marginal aqueous solubility (0.85 μM at pH 7) deterred its advanced assessment. Instead, we elected to focus initially on the monoaryl ethers R-6 and R-136, together with phenylpyridines R-84 and R-89, which all produced favorable mouse PK data, including excellent oral exposure levels and half-lives of 6−30 h (Table  6 and Figure S1). The selected candidates were further assessed in the chronic infection hamster model, which is considered the bona fide experimental model for VL because it mimics many features of progressive human disease. 51 Pleasingly, at 50 mg/kg twice daily for 5 days, phenylpyridine R-84 achieved 99.9−100% L. inf clearance in all three target organs, and R-89 was almost as good (Table 7 and (Table 6 and Figure S2)].
To better discriminate among the four preferred candidates, we next considered key safety features, starting with measuring their interactions with the hERG channel. While the monoaryl ethers R-6 and R-136 posed minimal risk (hERG IC 50 s of >30 μM), unfortunately, phenylpyridines R-84 and R-89 both caused potent inhibition (IC 50 s of 0.81 and 0.92 μM, respectively), indicating a strong likelihood of QT prolongation 52 (lead optimization criteria 53,54 mandate an IC 50 of >10 μM). This outcome was not anticipated, as 6S counterparts had generated much less concern. The two remaining compounds

Journal of Medicinal Chemistry
Article were then checked for any evidence of mutagenicity in the Ames test. Here, phenyl ether R-6 was negative, but the more soluble pyridinyl ether R-136 unexpectedly yielded a positive result. Although several other nitroimidazole drugs are Ames positive (e.g., metronidazole and fexinidazole), 55 this outcome effectively ruled R-136 out of contention because it would face a more difficult path to achieving regulatory approval. 16,53 Thus, R-6 was identified as the optimal VL lead. Further Appraisal of VL lead R-6. Additional properties of R-6 were measured and weighed against those of the initial preclinical candidate 4 ( Table 8). The two compounds were comparable in terms of molecular weight (345 Da vs 359 Da) and were both highly permeable, but R-6 had a lower measured LogD value (2.59 vs 3.10 for 4 and 2.52 for S-1 48 ), superior thermodynamic solubility (23 μM vs 2.8 μM), and a reduced propensity to bind to plasma proteins in various species (82− 87% vs 92−96% for 4). This lead also showed only weak CYP3A4 activity (IC 50 > 40 μM) and produced a notably favorable rat PK profile, with prolonged exposure and 100% oral bioavailability (Table 6 and Figure S2). These attributes reinforced our conclusion that R-6 (DNDI-8219) was indeed a very promising backup candidate for VL.
A larger scale synthesis of R-6 has recently provided a single 170 g batch of high-quality material (HPLC purity of >99.9% and 97.2% ee) in reasonable overall yield (8% over nine linear

Journal of Medicinal Chemistry
Article steps, starting from commercial R-solketal). However, the current synthetic route would still require significant improvement to deliver a scalable, robust, and cost-effective chemical process, in line with the stated TPP objectives for an affordable drug. Possible alternatives include using the known 56 enantiomer of orthogonally diprotected triol 125 and following the pathway described for S-6 in Scheme 4B because this enantiomer could be obtained from the cheap, optically pure starting material D-mannitol (via the less expensive Ssolketal). 57 Mechanistic studies of the closely related nitroimidazooxazine R-1 demonstrated that it was not activated by the previously identified type I nitroreductase (NTR1) in Leishmania, which mediates the cidal effects of monocyclic nitroheterocyclic drugs such as fexinidazole and nifurtimox. 38,58 Instead, the activity of R-1 was solely triggered by the same novel flavin mononucleotide-dependent NADH oxidoreductase (NTR2) in Leishmania that was employed by nitroimidazooxazoles, such as 4 and 5. 27 This new mode of action was elucidated through a combination of quantitative proteomics and whole genome sequencing of susceptible and drug resistant L. don promastigotes, the latter being generated via culture in the continuous presence of R-1 for 80 days (leading to a reduced level of expression of NTR2). 27 The further observation that R-1 and fexinidazole sulfone displayed additive effects against drug susceptible L. don sparked the suggestion of combination therapy between monocyclic and bicyclic nitro drugs to reduce the likelihood of any future clinical drug resistance. 38 However, we consider that it may be more preferable to look for alternative partner drugs for 3 (or R-6) with greater diversity in their mechanism of action.
One final aspect to consider with R-6 was its efficacy against a wider range of VL and cutaneous leishmaniasis (CL) strains. Overall, R-6 displayed potent broad-spectrum activity against both reference strains and clinical isolates (Table 9) These data confirm that R-6 has excellent potential as a therapy for VL and may have an additional application for the treatment of CL (the more common skin lesion form of leishmaniasis). 1 SAR of 6-Substituted 2-Nitroimidazo(or 2-Nitrotriazolo)oxazines for Chagas Disease. While the primary objective of this reinvestigation of pretomanid analogues was to develop a backup drug candidate for VL, retrospective screening against the protozoan parasites T. cruzi and T. brucei presented an opportunity to assess the possible capacity of these compounds to treat Chagas disease and human African trypanosomiasis (HAT), respectively. A brief inspection of Tables 1−4 found that only one compound (116) had submicromolar activity against HAT (T. brucei IC 50 of 0.63 μM), and this hit could be disregarded on the basis of its less favorable mouse PK profile, inferior aqueous solubility, and

■ CONCLUSIONS
In response to a compelling clinical need for more satisfactory VL treatments, recent efforts have been made to reposition leads from other therapeutic areas, seeking to accelerate new drug development. Promising results with antitubercular nitroimidazooxazoles and 7-substituted 2-nitroimidazooxazines encouraged us to evaluate additional scaffolds, e.g., nitroimidazooxazepines and methylated or reversed C-6 linker analogues of pretomanid, but these lacked sufficient potency and/or suitable PK and efficacy in the L. don mouse model. However, phenotypic screening of our pretomanid analogue library and follow-up IC 50 testing unveiled more active hits spanning a wide lipophilicity range (CLogP values of 1.2−4.5), including several with better solubility and microsomal stability, e.g., phenylpyridines and benzamide S-151. This work also pointed to the generally improved activities of novel 6R enantiomers, which was confirmed for phenylpyridines through comparative appraisal in the mouse VL model. Further studies in this series established that a 4-trifluoromethoxy phenyl substituent, para linkage, and a proximal 2-pyridine ring were preferred for good in vivo PK and efficacy, with two such leads (R-84 and R-89) giving ≥99% parasite clearance in the L. inf hamster model at 50 mg/kg b.i.d. These compounds also showed high potencies against T. cruzi, but unexpectedly high levels of hERG inhibition ultimately terminated their development.
Meanwhile, investigation of the C-6 linker group revealed that the parent ether moiety in R-1 (OCH 2 ) was nonoptimal for VL, with shorter and longer chain variants (O and OCH 2 CC, respectively) improving potency against L. inf (whereas O-carbamate and N-linked alternatives were poor). Although O-linked phenylpyridine R-168 displayed superb activity in the mouse VL model (99.7% reduction in parasite burden at 6.25 mg/kg, similar to the case for 4), we elected to focus instead on two less lipophilic monoaryl leads (R-6 and pyridine R-136), having superior solubility values (12−110 μg/ mL), low hERG risk, and excellent PK profiles in three species (mouse, rat, and hamster). Both compounds delivered high efficacies in the chronic infection hamster model (≥97% inhibition at 25 mg/kg, b.i.d.) and showed weakened binding to plasma proteins, although a positive Ames test for pyridine R-136 dissuaded its further advancement and earmarked R-6 as the favored VL backup candidate to 3. Finally, like the phenylpyridines, R-6 also demonstrated interesting activity against T. cruzi, whereas nitrotriazolooxazine congeners of such leads were less effective. These results provide new insights into the exciting potential of bicyclic nitroimidazoles as novel therapies for the treatment of some challenging neglected diseases.

■ EXPERIMENTAL SECTION
Combustion analyses were performed by the Campbell Microanalytical Laboratory, University of Otago, Dunedin, New Zealand. Melting points were determined using an Electrothermal IA9100 melting point apparatus and are as read. NMR spectra were measured on a Bruker Avance 400 spectrometer at 400 MHz for 1 H and 100 MHz for 13 C and were referenced to Me 4 Si or solvent resonances. Chemical shifts and coupling constants were recorded in units of parts per million and hertz, respectively. High-resolution electron impact (HREIMS), chemical ionization (HRCIMS), and fast atom bombardment (HRFABMS) mass spectra were recorded on a VG-70SE mass spectrometer at a nominal 5000 resolution. High-resolution electrospray ionization (HRESIMS) mass spectrometry was conducted on a Bruker micrOTOF-Q II mass spectrometer. Low-resolution atmospheric-pressure chemical ionization (APCI) mass spectra were obtained for organic solutions using a ThermoFinnigan Surveyor MSQ mass spectrometer connected to a Gilson autosampler. Optical rotations were measured on a Schmidt + Haensch Polartronic NH8 polarimeter. Column chromatography was performed on silica gel (Merck 230−400 mesh). Chromatographed compounds were typically further purified by crystallization from two solvent combinations, e.g., CH 2 Cl 2 and n-hexane, EtOAc and n-hexane, Et 2 O and n-pentane, or CH 2 Cl 2 and n-pentane (occasionally, Et 2 O was added to the latter combination to induce solidification, while some compounds required cooling at −20°C); more polar compounds were first dissolved in a minimum of 10% MeOH/CH 2 Cl 2 and slowly diluted with n-hexane to give the solid product. Thin-layer chromatography was performed on aluminum-backed silica gel plates (Merck 60 F 254 ), with visualization of components by UV light (254 nm), I 2 , or KMnO 4 staining. Tested compounds (including batches screened in vivo) were all ≥95% pure, as determined by combustion analysis (results within 0.4% of theoretical values) and/or by HPLC conducted on an Agilent 1100 system with diode array detection, using a 150 mm × 3.2 mm Altima 5 μm reversed phase C18 column or a 150 mm × 4.6 mm Zorbax Eclipse XDB 5 μm C8 column and eluting with a gradient (40 to 100%) of 80% CH 3 CN/water in 45 mM ammonium formate buffer (pH 3.5). Finally, the chiral purity of lead R-6 was assessed by HPLC performed on a Shimadzu 2010 system with diode array detection, employing a 150 mm × 4.6 mm CHIRALPAK AY-H 5 μm analytical column and isocratic elution with 20% EtOH/n-heptane.
Compounds of Table 1. The following section details the syntheses of compounds 14, 39, and 41 of Table 1, via representative procedures and key intermediates, as described in Scheme 1. For the syntheses of the other compounds in Table 1, see the Supporting Information.