7-Substituted 2-Nitro-5,6-dihydroimidazo[2,1-b][1,3]oxazines: Novel Antitubercular Agents Lead to a New Preclinical Candidate for Visceral Leishmaniasis

Within a backup program for the clinical investigational agent pretomanid (PA-824), scaffold hopping from delamanid inspired the discovery of a novel class of potent antitubercular agents that unexpectedly possessed notable utility against the kinetoplastid disease visceral leishmaniasis (VL). Following the identification of delamanid analogue DNDI-VL-2098 as a VL preclinical candidate, this structurally related 7-substituted 2-nitro-5,6-dihydroimidazo[2,1-b][1,3]oxazine class was further explored, seeking efficacious backup compounds with improved solubility and safety. Commencing with a biphenyl lead, bioisosteres formed by replacing one phenyl by pyridine or pyrimidine showed improved solubility and potency, whereas more hydrophilic side chains reduced VL activity. In a Leishmania donovani mouse model, two racemic phenylpyridines (71 and 93) were superior, with the former providing >99% inhibition at 12.5 mg/kg (b.i.d., orally) in the Leishmania infantum hamster model. Overall, the 7R enantiomer of 71 (79) displayed more optimal efficacy, pharmacokinetics, and safety, leading to its selection as the preferred development candidate.


■ INTRODUCTION
The neglected tropical disease visceral leishmaniasis (VL) is the second deadliest parasitic disorder (after malaria), being most prevalent in Brazil, Sudan, Ethiopia, and the Indian subcontinent, with an estimated 350 million people at risk of infection. 1 Transmitted by sand flies, the disease first manifests as an irregular fever, anemia, leukopenia, and hepatosplenomegaly and is usually fatal within two years if left untreated. 2 About 300000 new cases arise annually, almost half in children, and at least 35 countries have reported the occurrence of HIV coinfection (with up to 34% incidence), which gives a significantly higher mortality rate. 3,4 Unfortunately, none of the existing VL drugs (antimonials, paromomycin, liposomal amphotericin B, or miltefosine 1; see Figure 1) is universally effective nor free from further drawbacks such as parenteral administration (for all except 1), toxicity, high cost, and emerging resistance. 5 Furthermore, there is no available vaccine despite renewed efforts. 6 Clinical investigation of the orally active aminoquinoline sitamaquine (2) has been abandoned due to its toxicity and less satisfactory efficacy, 7 and new phase II trials of the repositioned oral agent fexinidazole (3) 8 for VL have also been interrupted due to patient relapses. 9 With no other candidates under clinical evaluation at present, there is a desperate need for the development of more effective, safe, and affordable oral remedies for VL.
We have recently reported that phenotypic screening by the Drugs for Neglected Diseases initiative (DNDi) of some nitroimidazole derivatives arising from our early studies with the TB Alliance unexpectedly led to the identification of DNDI-VL-2098 (4) as a preclinical candidate for VL. 10,11 Our opening assignment with TB Alliance had been to prepare and evaluate novel nitroheterobicyclic analogues of the tuberculosis (TB) drugs delamanid (5) and pretomanid (PA-824, 6), 12,13 seeking a possible third active scaffold for the construction of a backup series. However, among the fused 5/6 ring systems examined, only the metabolically labile 2-nitroimidazothiazines retained significant antitubercular potency, 14 returning our attention to the original oxazine class where we uncovered heterobiaryl derivatives of 6 with better efficacy (e.g., TBA-354, 7). 15,16 One important consideration in the design of a superior secondgeneration TB candidate was the potential for cleavage of the aromatic side chain via oxidative metabolism of the 6oxymethylene linker; therefore, several alternative linker and steric protection strategies were explored, albeit with limited success. 17−19 A final, more innovative way to address this issue was to invoke a scaffold hopping approach 20 by relocating aromatic side chains from the 6-position to the 7-position of the 2-nitroimidazooxazine core, with attachment via the same inverted linker (CH 2 OR) that was present in 6-nitroimidazooxazole 5. This was equivalent to a one carbon expansion of the oxazole ring between C-2 and C-3 ( Figure  2). The rationale for this design concept stemmed from initial evidence 21 that delamanid (5) was highly stable toward metabolism as well as from a report 22 that 7-methyl derivatives of 6 retained excellent antitubercular potency, suggesting that such an approach merited investigation.
Serendipitously, we soon discovered 23 that this novel "7substituted oxazine" class not only showed considerable promise for TB (as later confirmed by others 24,25 ), it also displayed potent antileishmanial activity comparable to the 6nitroimidazooxazoles in early screening assays. Therefore, following the success with 4, this new series was similarly repositioned for VL as part of an extensive backup program run in collaboration with DNDi. In this paper, we first highlight some critical VL hit to lead assessments on the original subset of compounds that had been prepared for TB. We then detail the findings of our lead optimization study directed at developing backups to 4 having an improved physicochemical/pharmacological profile and better safety, which culminated in the selection of a new preclinical candidate for VL. Finally, in light of these encouraging results and the excellent activities of this novel 7-substituted 2-nitroimidazooxazine class against both TB and Chagas disease, we point to related analogues that might be worthy of further assessment for the latter applications.

Journal of Medicinal Chemistry
Article ring closed to 7-substituted oxazines, as above. An attempt to combine the last two steps in one pot 30 (by exposing 42 to 1.2 equiv of NaH and 4-(trifluoromethoxy)phenol in DMF at 75− 86°C) led to markedly inferior results (27% 44, with 30% 43); equally, ring opening of 42 with 4-iodophenol in DMF (K 2 CO 3 , 83°C, 8 h) also gave a lower yield of 50 (60%) due to partial displacement of the 2-chlorine. Suzuki couplings on the ring-closed iodide 51 readily furnished biphenyl derivatives 52 and 54; terminal fluoropyridines 55 and 56 were similarly obtained from 48 and 51 through the use of a weaker base (KHCO 3 ).
For more efficient synthesis of 7-H biaryl analogues having a proximal 3-pyridine ring, an epoxide-opening strategy (Scheme 3A) was preferred over the Mitsunobu route described above.
By alkylating 2-chloro-4-nitroimidazole (40) with (4R)-4-(2iodoethyl)-2,2-dimethyl-1,3-dioxolane 33 (72) (or its optical isomer, 80 33 ), it was possible to transform the above racemic route into a viable chiral synthesis for delivery of both enantiomers of two advanced leads, 71 and 93 (Scheme 3B,C). The two chiral acetal products 73 and 81 were readily converted into the R and S enantiomers of epoxide 67 (76 and 84) by successive hydrolysis (to diols 74 and 82), tosylation at the primary hydroxyl, and internal substitution to form the oxirane ring (DBU). These chiral epoxides were then elaborated to the final products by reaction with 68, ring closure, and Suzuki coupling, as previously described.
The preparation of biaryl congeners 123, 126, 129, 131− 133, and 139 in which the first ring was pyridazine, pyrazine, or pyrimidine followed similar procedures to those developed for the pyridine analogues. Thus, sodium hydride-induced S N Ar reactions of alcohols 13 and 20 with haloheterocycles 121, 124, and 127 readily provided the bromoheteroaryl ether intermediates needed for final step Suzuki couplings (Scheme 4A). However, the remaining arylpyrimidine target (139) required prior assembly of the biaryl side chain. Initial protection of 2-chloropyrimidin-5-ol (134) as an ethoxymethyl ether derivative (135), followed by Suzuki coupling and acidic deprotection, supplied arylpyrimidinol 137 in excellent yield (83% from 134; Scheme 4B). Reaction of 137 with epoxide 67 produced a 5:2 mixture of the alcohol 138 and the ring closed oxazine (139); treatment of 138 with sodium hydride then completed this synthesis.
Scheme 5 outlines the methods used to obtain compounds 142, 144, 147, 149, 152, 154, 157, 159−161, 170, and 178, whose side chains contained either piperazine or piperidine linked to an aryl group. Ring opening of epoxides 67 and 42 with the known or commercial amines 140, 145, 150, 34 and 155 easily generated the expected β-amino alcohols in high yield (Scheme 5A,B). These alcohols could be ring closed to the final products with sodium hydride upon mild heating, albeit yields for the 7-methyl analogues were generally significantly lower, in part due to greater purification difficulties. Chloroformylation of alcohols 13 and 20 (triphosgene/Et 3 N) and in situ reaction with arylpiperazine 140 also led to the Ocarbamates 160 and 161 in only modest yield (33−35%; Scheme 5C) on account of similar purification issues; the isolation of alkyl chloride and diethyl carbamate derivatives under the same reaction conditions has been reported recently. 35 Lastly, synthesis of the two O-linked arylpiperidines, 170 and 178, was eventually achieved in each case via a lengthy seven-step route (Scheme 5D), after the failure of a more direct plan (ring opening of epoxide 67 with piperidinol 163 36 in the presence of erbium triflate 37 ). Here, piperidinol 163 36 was first sourced in three steps by Buchwald amination of 1-bromo-4fluorobenzene with 1,4-dioxa-8-azaspiro [4.5]decane, 38 ketal hydrolysis, and reduction (NaBH 4 ). Reaction of epoxides 162 39 and 171 39 with 163 (NaH, DMF, 70°C) and TBS protection of the liberated hydroxyls provided the desired ethers 165 and 173 in good yield (47−68% overall). Successive benzyl removal (via hydrogenolysis), iodination, and alkylation of 2-chloro-4-nitroimidazole (40) then gave the TBS-protected adducts, 168 and 176, which were readily desilylated (TBAF) and ring closed to furnish the final targets.

■ RESULTS AND DISCUSSION
The structures and in vitro antiparasitic and antitubercular potencies of 75 novel 7-substituted 2-nitroimidazooxazine derivatives prepared in two collaborative projects are provided in Tables 1 and 2. While compounds 14, 21−23, 25, 27−29, 34, 38, 39, 44, 45, 47, 49, and 52−54 were initially designed and evaluated for TB, for clarity purposes, we will focus the discussion first on the more recent VL work with DNDi. Here, new synthesis was directed at the optimization of solubility, efficacy, and safety, primarily through the incorporation of heterocycles to reduce compound lipophilicity 11

Journal of Medicinal Chemistry
Article Canada). Kinetic aqueous solubility measurements were conducted on dry powder forms of particular examples that were being considered for further evaluation. Target compounds were initially screened only once against Leishmania donovani (L. don) using a mouse macrophage-based luciferase assay conducted at the Central Drug Research Institute (CDRI, India). 10 Nevertheless, to gain a clearer understanding of the SAR (in view of some unexpected in vivo outcomes), the entire set was finally re-evaluated at the University of Antwerp (LMPH) in replicate assays against three protozoan parasites: Leishmania infantum (L. inf), Trypanosoma cruzi, and Trypanosoma brucei. 40 Assessments of cytotoxicity were concurrently conducted on both human lung fibroblasts (MRC-5 cells; the host for T. cruzi) and primary peritoneal mouse macrophages (the host for L. inf), which revealed that the compounds were generally nontoxic (MRC-5 IC 50 s > 55 μM except for 117: IC 50 35 μM), as confirmed for TB (VERO assay 41 IC 50 s > 128 μM for 71 of 72 compounds).
Early Hit to Lead Assessments for VL. Through an agreement between TB Alliance and DNDi, 58 nitroimidazole derivatives were screened against L. don at the Swiss Tropical Institute. All five 7-substituted oxazines (including 22, 23, and 28) demonstrated excellent potencies in the in vitro mouse macrophage assay (IC 50 s 0.065−0.17 μM, similar to racemic 4), prompting the inclusion of 28 alongside rac-4 (and another oxazole 11 ) in a proof-of-concept in vivo assessment at the London School of Hygiene and Tropical Medicine (LSHTM). However, the level of activity observed for 28 in this L. don mouse model (49% inhibition at 50 mg/kg, dosing po daily for 5 d; Table 3) was not notable in comparison to the results for rac-4 (99% at 6.25 mg/kg), 11 suggesting that further optimization of the side chain would be necessary. Indeed, while 28 showed good stability on exposure to mouse liver microsomes (MLM: 75% remaining after 1 h, Table 3) and gave a mouse pharmacokinetic (PK) profile comparable to rac-4 (Table 4 and Supporting Information, Figures S1 and S2), it was very hydrophobic (CLogP 5.14) and displayed poor solubility (∼58 ng/mL at pH 7, Table 3; 62-fold lower than for rac-4 11 ). Such compounds typically exhibit high levels of plasma protein binding (PPB), which can limit efficacy. 42 We have also observed that increased linker flexibility can be detrimental to in vivo activity. 17,43 While these mouse studies were being conducted, a further 30 7-substituted oxazine derivatives were screened against L.

Journal of Medicinal Chemistry
Article don in the luciferase assay at CDRI. 10 On the basis of the single IC 50 data obtained for 14, 21 −23, 25, 27−29, 34, 38, 39, 44, 45, 47, 49, and 52−54 (Table 1), several preliminary SAR conclusions were drawn: (1) the 7-H series was generally 5− 10-fold more potent than the 7-methyl series; (2) 4trifluoromethoxy and 4-benzyloxy substituents (forms A and C) provided equivalent potency; (3) for biaryl analogues (forms B and D), 4-fluoro was preferred over 4-trifluoromethoxy as the final ring substituent (as observed 11 in the 6nitroimidazooxazole series); (4) a shorter linker (forms C and D) was preferred in the majority of cases. Thus, the most active analogues appeared to be 14,22,25,39,45,49, and 53 (IC 50 s 0.01−0.06 μM, similar to 4). However, benzyl ether 14 did not display suitable metabolic stability (10% parent left after 1 h with MLM; Table 3), while evidence from the 6-substituted oxazine series 17 for the more rapid metabolism of benzyloxybenzyl analogues dissuaded further testing of 22 and 45. Moreover, following the disappointing results with 28, we were not optimistic of good in vivo efficacy with close analogue 25 despite its improved potency. Therefore, we elected to initially investigate 39, 49, and 53 as potential leads, together with two counterparts from the 7-methyl series, 44 and 54, to enable a head-to-head comparison.
The selected compounds were advanced to parallel mouse PK profiling and efficacy studies in the mouse VL model. Encouragingly, both phenyl ether 44 (the direct analogue of rac-4) and biphenyl congener 49 showed excellent efficacy at 25 mg/kg (99.9−100% inhibition; Table 3 and Figure 3b). Surprisingly, the more potent 7-H counterpart of 44 (39) was slightly less active in this assay (87% inhibition), mimicking findings for the 2-H equivalent of rac-4. 11 Moreover, the biphenyl derivatives of 39 and 44 (53 and 54) were also less impressive than 49 (65% and 30% inhibition, respectively). However, while these latter results appeared to track well with the single determination L. don data, they did not seem to line up with the almost equivalent mean potencies vs L. inf ( Table  1). The findings also appeared to conflict with the kinetic solubility and microsomal stability data (Table 3), where 49 was as poorly soluble as 28 (55 vs 58 ng/mL), but the more stable analogue 54 (85 vs 75% in MLM) was 45-fold more soluble than 28 (2.6 μg/mL). Solubility is discussed further in the next section.
Analysis of the mouse PK data ( Table 4) provided greater insight, revealing that 39 had a 4-fold higher rate of clearance than its 7-methyl derivative 44 (48 vs 12 mL/min/kg), resulting in a short half-life (1.1 h vs 2.8 h for 44) and quite poor oral exposure (see the Supporting Information, Figure  S1). Interestingly, with iv administration, the PK profiles for 44 and rac-4 were fairly similar, but 44 did not perform as well under oral dosing, with rather modest absorption (C max 1.4 μg/ mL, 3-fold less than for rac-4) contributing to reduced exposure and moderate oral bioavailability (35% vs 79%). The oral parameters for compound 54 were also mediocre (poor C max of 0.79 μg/mL and low oral bioavailability of 17% offsetting its lengthy 27 h half-life), potentially explaining its inferior efficacy in the mouse VL model. However, the findings for 49 and 53 were more puzzling, with the less efficacious 53 demonstrating greater oral exposure (see the Supporting Information, Figure  S1), superior oral bioavailability (100% vs 11% for 49), and an extended half-life (17 h vs 6.7 h for 49). Nevertheless, like 28, 53 was particularly hydrophobic (CLogP 5.03), so high PPB may be a major issue limiting its efficacy. 42 We have previously observed that PK data is not always correctly predictive of in vivo efficacy ranking. 43 These promising results prompted further appraisal of the most active compounds, 44 and 49. In the mouse VL model, 44 provided robust dose−response data (Table 3), giving an ED 50 value of 4.2 mg/kg (cf. 3.0 mg/kg for rac-4 11 ). Unfortunately, additional studies of 49 in this model (using material prepared elsewhere) were unable to replicate the original result; we postulate that this discrepancy may be due in part to the extremely poor aqueous solubility and inadequate oral bioavailability of this compound, rendering oral suspension formulations particularly sensitive to particle size. Nevertheless, the optimal in vivo assay for assessing the efficacy of test compounds against VL is the chronic infection hamster model, which better reproduces the clinical pathology of human disease. 44 In the L. don hamster model at CDRI, leads 44 and 49 were almost equally effective at 50 mg/kg, with 5 days of oral dosing leading to 53% and 51% inhibition of parasite infection in the spleen, whereas rac-4 gave 86% inhibition under the same dose regimen. 11 A significant factor in the suboptimal activity of 44 in the hamster model was thought to be its exceptionally rapid metabolism in this species, as revealed by the hamster microsomal stability data (only 16% remaining after 0.5 h vs 49% for rac-4 11 ). Therefore, 44 was later reassessed in the L. inf early curative hamster model at LMPH, comparing a twice-daily dose regimen (25 mg/kg b.i.d.) with a once daily dose of 50 mg/kg. The results (Table 5) slightly favored the twice daily regimen for all three target organs; hence, this protocol became standard for most test compounds. However, unlike 4, 44 was not curative at this dose level. Another liability with 44 was its greater inhibition of the hERG channel (IC 50 3.8 vs 10.5 μM for 4), with IC 50 values in excess of 10 μM required to minimize QT prolongation risk. 45 Hence, as lead compounds for VL, 44 and 49 fulfilled many suggested criteria 46 but still had key deficiencies, reflecting their origin as screening hits in a scarcely studied new class. SAR of 7-Substituted 2-Nitroimidazooxazines for VL. Following the identification of 4 as a preferred drug candidate and the discovery of 44 and 49 as unoptimized new leads, a    41 or hypoxic (LORA) 56 conditions.

Journal of Medicinal Chemistry
Article backup program was launched to develop second-generation agents for VL having better solubility, PK−PD, and safety profile. 11 Because of the inferior profile of 44 in comparison to 4 in several key areas, we elected to center our synthetic strategy mainly on bicyclic side chains, employing heterocycles to modulate lipophilicity and solubility. Six-membered ring nitrogen-containing variants were preferred due to their greater metabolic stability; 47 ortho-substitution of aryl groups and metalinkage of rings were also investigated as additional options to increase solubility. 48 Recognizing that few orally active registered drugs have solubility values below 1 μM at pH 7.4 (the pH of blood), 49 we aspired to achieve at least 10-fold higher than this for the best compounds. 46 We also aimed to exploit the low pH of gastric fluid (∼1−2) to improve dissolution and oral absorption of analogues containing pyridine and other bases. 50 Hence, we set a minimum solubility requirement for the preferred final candidate of being noninferior to delamanid (5) (0.31 μg/mL at pH 7 and 116 μg/mL at pH 1), 11 an approved TB drug in Europe and Japan. 12 On the basis of the wider in vitro screening results, it was apparent that the 7-substituted oxazines could not be used for African trypanosomiasis (T. brucei IC 50 s mostly >64 μM, none <1 μM; see Supporting Information, Tables S1 and S2). However, unlike the 6-nitroimidazooxazoles, this new oxazine class generally showed interesting potencies against T. cruzi (IC 50 s 0.03−1 μM), suggesting the possibility of dual utility to treat both VL and Chagas disease. Further analysis of data for the 65 racemic compounds tested indicated a modest trend for the best VL leads to have high potencies against T. cruzi (see Supporting Information, Figure S4). Hence, for simplicity, we will focus this part of the SAR discussion entirely on the intended primary application (VL), emphasizing the key L. inf results.

Article
To begin with, a reanalysis of the initial data set (up to and including 54; Table 1) confirmed weak trends on L. inf for the 7-H analogues to be more potent and a shorter linker length to be preferred (e.g., 39: IC 50 0.047 μM), but there was no consistent preference for 4-fluoro as the terminal ring substituent. Nevertheless, in view of the better in vivo efficacy of 49 and similarly substituted nitroimidazooxazoles, 11 we retained this latter design element in the majority of cases. Thus, compounds 55 and 56 first investigated the effect of replacing the second phenyl ring of 49 by pyridine (ΔCLogP −1.2 units). Pleasingly, this led to a 2−6-fold potency increase, with 55 (IC 50 0.083 μM) also being 6.5-fold more soluble than 49 (0.36 vs 0.055 μg/mL, Table 3). Exchange of the first phenyl ring by 2-pyridine (59 and 61; ΔCLogP −0.5 units) resulted in even better activity (59: IC 50 0.050 μM), and in this case solubility values were ∼20-fold higher at low pH (2.8−13 μg/mL; calcd pK a 2.83) although still rather modest. Therefore, we examined the addition of an ortho fluorine in the phenyl ring (62 and 63) in an attempt to break up the planarity. 48 However, while this change was well tolerated, there was no improvement in solubility and microsomal stability was reduced (19% vs 43% in MLM for 62 vs 59, Table 3). In an alternative approach, we tried meta-linkage of the rings (99 and 101−103), but although the activity was generally acceptable, this led to inferior solubility (99: 27 ng/mL).

Journal of Medicinal Chemistry
Article (79 and 94) as slightly preferred for both potency and microsomal stability (particularly in the case of 79).
In view of the promising results with phenylpyridines, we elected to investigate the more hydrophilic bipyridines (105− 120). Most of these showed interesting potencies in the initial L. don screen and further assessments had identified 108, 112, and 113 as being of potential interest based on their improved solubilities in comparison to 49 (2.3−4.5 vs 0.055 μg/mL). However, on retesting, almost all of the 7-H compounds displayed markedly inferior utility against L. inf (IC 50 s 2.5 to >64 μM) while the 7-methyl bipyridines retained moderate potencies (IC 50 s 0.20−1.1 μM). It is intriguing to speculate that this might indicate a "minimum lipophilicity" requirement for activity (e.g., CLogP ∼ 2.5) because a similar pattern was noted for all of the more hydrophilic analogues (see analysis of racemic 7-H data set, Figure 4). Another strategy for heterobiaryl analogues of 49 was to exchange the first phenyl ring with pyridazine, pyrazine, or pyrimidine (123, 126, 129, 131−133, and 139). Of these, pyrimidine (129, 131, and 139) provided the best activity (IC 50 s 0.21−0.29 μM), although combining this with a pyridine ring (132, 133) led to a dramatic loss of potency (21-to >220-fold). Overall, pyrimidine 129 had the best aqueous solubility (1.8 μg/mL; 33-fold better than 49) along with acceptable metabolic stability.
More structurally diverse targets (142, 144, 147, 149, 152,  154, 157, 159−161, 170, and 178; Table 2) were designed on the premise that arylated cyclic amines can be effective bioisosteres for biphenyls, thus facilitating substantial boosts in solubility. 51,52 Several side chains of this type have previously shown promise for TB and/or VL, 11,18 including in the recent development of antileishmanial aminopyrazole ureas. 53  Integration of the initial L. don data with the kinetic solubility and microsomal stability results led to the selection of nine new racemic analogues of 49 for testing in the L. don mouse model (dosing at 50 mg/kg for 5 d; Table 3 and Figure 3a). Encouragingly, a first experiment on 3-pyridine derivative 71 (4-FPh) yielded a 100% parasite clearance from the liver in all mice. Following this, 4-trifluoromethoxy congener 93 was found to be equally efficacious (99.5%), whereas the 2,4difluoro example 91 was slightly less effective (91% inhibition). However, the less potent 7-methyl derivative of 71 (90) and the more potent 2-pyridine analogue 59 were only moderately active (41% and 67%, respectively); it is possible that the higher crystallinity (larger particle size) of 59 may have contributed to poor oral bioavailability. 42,48 Two more soluble heterobiaryl analogues, bipyridine 112 and phenylpyrimidine 129, also displayed lower efficacy (44% and 85% inhibition); oral PK data on 112 (Table 4 and Supporting Information, Figure S2) were comparable to those of 71, so this may be a potency issue (as suggested by the disparate L. inf and L. don IC 50 s of >64 vs 0.09 μM). Finally, the inferior in vivo outcomes for two potential bioisosteres of 49, phenylpiperazine 142 (55%) and O-linked phenylpiperidine 170 (45%), may be attributed to either weaker in vitro activity on retesting (for 142: L. inf IC 50 2.3 μM) or more rapid metabolism (for 170), as indicated above. No adverse effects were noted in any of the in vivo experiments and the percentage weight changes for the mice were well within normal thresholds (see the Supporting Information, Table S4).
Dose−response appraisal of 71 in this mouse model provided an ED 50 value of 5.1 mg/kg (cf. 4.2 mg/kg for 44), whereas the trifluoromethoxy analogue 93 was unexpectedly

Journal of Medicinal Chemistry
Article ∼3-fold better (50% at 1.56 mg/kg;  Figure 5), similar to 4 at 25 mg/kg once daily (qd). However, 71 was slightly less effective than 4 when given via the 12.5 mg/kg qd schedule. A final head-to-head comparison of the enantiomers of 71 confirmed 79 as the preferred stereoisomer based on its superior efficacy at two dose levels. This result was also supported by favorable PK data, e.g., a higher exposure than 87 in hamsters, with an acceptable half-life (3.1 h) and good oral bioavailability (34%) in the rat (Table 4 and Supporting Information, Figure S3).
Although 94 was not tested in the hamster model, it is thought that 79 may still offer some advantages as a lead candidate, e.g., lower lipophilicity (by ∼1 log unit) and reduced molecular weight (this could lessen PPB and improve safety), 42 slightly better solubility (a calculated pK a value of 3.76 vs 3.42 for 94), and a physical form more suitable for oral administration.
In line with our initial objective to develop improved drug candidates as backups to 4, it was pertinent to examine some additional properties of 79 (Table 6). Compared to 4, 79 had a very similar molecular weight (370 vs 359 Da) and provided thermodynamic solubility values that were clearly superior to 4 as the pH approached the measured pK a value of 3.95. It also had a lower experimental Log D value (2.45 vs 3.10), close to that of pretomanid (6). 18 Furthermore, like 4, 54 79 displayed high permeability (without being a substrate for P-gp mediated efflux), although it did show a slightly greater binding to human plasma proteins (96.5 vs 93.9%). In terms of safety, 79 gave a low inhibition of hERG (IC 50 > 30 μM), did not inhibit CYP3A4 (IC 50 > 100 μM), and was not mutagenic (Ames test). These characteristics broadly match the suggested criteria for clinical development of a new entity for VL, 55 so following a belated concern with 4, 79 has now been selected as a new preclinical candidate. SAR of 7-Substituted 2-Nitroimidazooxazines for TB. Although the primary goal of our work with DNDi was a new drug for VL, the series was originally designed and exemplified for TB, seeking a novel second-generation backup to 6 (now in phase II/III clinical trials 13 ). Hence, the antitubercular activities of the 7-substituted oxazine derivatives have remained an aspect of significant ongoing interest. The work began with the preparation of an exploratory set of four compounds (14,22,39, and 45; Table 1). Growth inhibitory effects against Mycobacterium tuberculosis (M. tb, strain H37Rv) were studied under both aerobic (replicating) and hypoxic (nonreplicating) conditions (MABA 41 and LORA 56 assays, respectively), in recognition of the varying modes of action of 6 under each state 57 and the suggestion that optimizing for hypoxic activity may lead to agents with better sterilizing ability against persistent bacteria; 56 recorded MIC data (for at least 90% inhibition) represent the mean of 2−5 independent measurements. Compared with racemic 6 (MICs of 1.1 and 4.4 μM in MABA and LORA, respectively), 14 (Table 1). These featured two design elements that had proven most advantageous for enhancing in vivo efficacy in early studies of 4 and 6, namely biaryl extension, and methylation adjacent to the ring oxygen. 11,30,58 From this larger data set, it was observed that 7-methyl congeners (e.g., 21, 23, 44, and 47) were generally slightly more effective than 7-H counterparts and that biphenyl side chains (e.g., 25, 27−29, 49, and 52−54) provided roughly an order of magnitude further improvement in MABA MIC values (whereas LORA data were less responsive to these changes). The phenylbenzyl derivative 29 was earmarked as a potential

Journal of Medicinal Chemistry
Article early lead based on its better MIC profile (0.093 and 1.4 μM in MABA and LORA) and good stability toward MLM and HLM (77−85% parent remaining after a 1 h exposure, Table 3). Thus, for preliminary proof of principle, the enantiomers of 29 (34 and 38) were prepared and assessed in the acute TB infection mouse model alongside 6, dosing orally at 100 mg/kg daily (5 days/week) for three weeks. In this experiment, the R enantiomer 34 displayed equivalent efficacy to 6, but the S form 38 was 5-fold less active (Figure 6), in accordance with its weaker potency and MLM stability data. Nevertheless, the very high lipophilicity of 34 (CLogP 5.52) and its inferior PK profile in comparison to the shorter linked analogue 53 (Table 4) imply that far better in vivo effects might be achievable with optimized compounds (as shown for VL; cf. 94 vs 28).
The early preference for 7-methyl substitution was not a consistent pattern across heterobiaryl derivatives, where the most potent examples, notably phenylpyridines 59, 64, 93 and 94, as well as phenylpyrimidine 129 (MABA MICs 0.02−0.04 μM), were 7-H compounds. As found for the 6-substituted series, 16 bipyridine and other heterobiaryl analogues were generally less impressive (except the 4-CF 3 congener 111), particularly when the rings were meta-linked (corresponding phenylpyridines 99 and 101−103 also displayed markedly reduced activity). Finally, arylated cyclic amine bioisosteres ( Table 2) showed moderate to weak potencies overall, with the hydrophilic benzoylpiperazines 157 and 159 being especially poor. For side chains A−C, 7-methyl compounds exhibited an order of magnitude better aerobic activity than their 7-H counterparts, although LORA results were disappointing for these and the related O-linked phenylpiperidines (170 and 178). Nevertheless, it was recognized that most of these new 7substituted oxazines possessed a 4-fluoro substituted terminal ring, whereas more lipophilic 4-trifluoromethoxyphenyl (or 4-CF 3 pyridine) termini were favored for TB 16 Table S3) were also made and evaluated. Here, 191 and 198 (4-OCF 3 ) were 10-to 16-fold more effective than 123 and 139 (4-F) in both TB assays (192−194 were also 2−8-fold better than 126, 129, and 131 in MABA), confirming this same SAR pattern. Overall, taking into account potency 46 (needing to be superior to 5 11 ), solubility, and metabolism effects, it is considered that phenylpyridine 94 is the most promising lead for TB.

■ CONCLUSIONS
Through a scaffold hopping design strategy, 7-substituted 2nitroimidazooxazines were identified as a third, highly active nitroimidazole-based class of antitubercular agents, having a remarkable similarity in properties to 2-substituted 6-nitroimidazooxazoles. Phenotypic screening of some unoptimized early examples against kinetoplastid diseases led to the detection of two compounds (44 and 49) having significant efficacy against VL in mouse and hamster models, although these proved to be inferior to preclinical lead 4 as potential drug candidates. On the basis of our experiences in the original two classes (with 4 and 6), we then sought to develop more suitable second-generation agents for VL by systematically exploring heterocyclic side chain variants of biphenyl lead 49. Replacement of one or both phenyl rings by pyridine (or pyrimidine etc.) enabled large modulations in lipophilicity (ΔCLogP −0.5 to −2.7 units), with concomitant improvements in aqueous solubility (2−71-fold at pH 7 and ∼4000-fold at pH 1 for phenyl-3-pyridines). In a complementary bioisostere approach, the incorporation of piperazine or piperidine for the first ring produced even greater solubility enhancements (e.g., 170: 34 mg/mL at low pH). However, more subtle strategies (viz. ortho-substitution of aryl groups and meta-linkage of aryl rings) proved less beneficial overall.
Interestingly, potency against L. inf appeared to show some dependence on lipophilicity, with the most effective 7-H compounds falling in a CLogP range of 2.9−4.0, and compounds of CLogP < 2.5 having weak or negligible activity. This was aptly demonstrated by the improved potency of fluorinated phenylpyridines (5−16-fold over 49), in which the pyridine could be either terminal or proximal to the linker, whereas the combination of two pyridine rings was strongly deactivating except in the presence of a 7-methyl substituent. Phenylpyrimidine and phenylpiperidine were the only other side chains to provide substantial activity in this assay. It has recently been shown 59 that a novel nitroreductase (NTR2) in Leishmania is responsible for the activation of nitroimidazooxazoles such as 4; therefore, differential nitroreductase binding may be a major factor behind the in vitro SARs for both VL and TB. 57 Evaluation of a representative set of nine racemic compounds in the VL mouse model pinpointed phenylpyridines 71 and 93 as the most efficacious, with 93 being as impressive as 4 (50% inhibition at 1.56 mg/kg). In the chronic infection L. inf hamster model, 71 (at 12.5 mg/kg b.i.d.) achieved >99% reductions in parasite burden for all three target organs. Subsequent synthesis and assessment of the enantiomers of both leads identified the R forms (79 and 94) as superior, and in the case of 79 this outcome was reinforced by excellent results in the L. inf hamster model and favorable PK data in the hamster and rat. Importantly, 79 (DNDI-0690) also provided a better safety profile than 4 and has now been selected as a new preclinical candidate for VL.
Finally, as found for the nitroimidazooxazole series, 11 it was intriguing to note that some of the best VL leads (e.g., 59, 79, 93, 94, and 129) showed highly potent in vitro effects against TB, with both R enantiomers and 4-trifluoromethoxy analogues most preferred, pointing to 94 (MABA MIC 0.024 μM) as the favored TB candidate for further evaluation. The S form of 93 (95) also displayed interesting activity against T. cruzi (IC 50 0.13 μM), indicating a possible application for treating Chagas disease. This investigation has therefore revealed that the 7substituted 2-nitroimidazooxazine class has exciting potential to

Journal of Medicinal Chemistry
Article treat up to three neglected diseases and can deliver drug candidates that are worthy of examination in ongoing studies.

■ 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 ppm and hertz, respectively. High-resolution fast atom bombardment (HRFABMS) mass spectra were determined on a VG-70SE mass spectrometer at nominal 5000 resolution. High-resolution electrospray ionization (HRESIMS) mass spectra were determined 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). Thin-layer chromatography was carried out on aluminumbacked 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 ≥95% pure, as determined by combustion analysis (results within 0.4% of theoretical values) and/or by HPLC conducted on an Agilent 1100 system, using a 150 mm × 3.2 mm Altima 5 μm reversed phase C18 column with diode array detection. Preparative reversed phase HPLC was performed using a Gilson Unipoint system (322-H pump, 156 UV/ vis detector) with 250 mm × 21 mm Synergi Max-RP 4 μm C12 or Zorbax 7 μm SB-C18 columns. Finally, preparative chiral HPLC was carried out on similar equipment by employing a 250 mm × 20 mm CHIRALPAK IA 5 μm semipreparative column, while chiral purity was assessed using 250 mm × 4.6 mm CHIRALPAK IA or CHIRALPAK AS-H 5 μm analytical columns.
Compounds of Table 1. The following section details the syntheses of compounds 14, 25, 34, 44, 55, 59, and 79 of Table 1, via representative procedures and key intermediates, as described in Schemes 1−3. For the syntheses of all of the other compounds in Table 1, please refer to the Supporting Information.

Journal of Medicinal Chemistry
Article pharmacodynamic; M. tb, Mycobacterium tuberculosis; HRFABMS, high resolution fast atom bombardment mass spectrometry; HRESIMS, high resolution electrospray ionization mass spectrometry; APCI MS, atmospheric pressure chemical ionization mass spectrometry; HLM, human liver microsomes; CFU, colony forming unit; SD, standard deviation