Assessment of a pretomanid analogue library for African trypanosomiasis: Hit-to-lead studies on 6-substituted 2-nitro-6,7-dihydro-5H-imidazo[2,1-b][1,3]thiazine 8-oxides

Graphical abstract


a b s t r a c t
A 900 compound nitroimidazole-based library derived from our pretomanid backup program with TB Alliance was screened for utility against human African trypanosomiasis (HAT) by the Drugs for Neglected Diseases initiative. Potent hits included 2-nitro-6,7-dihydro-5H-imidazo [2,1-b] [1,3]thiazine 8-oxides, which surprisingly displayed good metabolic stability and excellent cell permeability. Following comprehensive mouse pharmacokinetic assessments on four hits and determination of the most active chiral form, a thiazine oxide counterpart of pretomanid (24) was identified as the best lead. With once daily oral dosing, this compound delivered complete cures in an acute infection mouse model of HAT and increased survival times in a stage 2 model, implying the need for more prolonged CNS exposure. In preliminary SAR findings, antitrypanosomal activity was reduced by removal of the benzylic methylene but enhanced through a phenylpyridine-based side chain, providing important direction for future studies Ó 2017 Elsevier Ltd. All rights reserved.
Human African trypanosomiasis (HAT, also known as sleeping sickness) is a particularly lethal neglected tropical disease that is endemic in remote sub-Saharan Africa. 1 HAT arises from infection by two subspecies of the kinetoplastid parasite Trypanosoma brucei (T. b. gambiense and T. b. rhodesiense), which are transmitted through the bite of tsetse flies. 2 Because symptoms of the initial bloodstream stage are fairly mild and non-specific (e.g., headache, fever, weakness), the disease often progresses to the potentially fatal CNS stage characterised by neurological and psychiatric disorders before treatment is sought. 1,3 However, there are pitifully few available drugs for late stage HAT, and all require hospitalization. 3,4 The antiquated first-line remedy melarsoprol (1, see Fig. 1) is highly toxic, causing death in $5% of patients, and is increasingly less effective due to drug resistance. 4,5 Eflornithine (2) is less toxic but more costly and cumbersome to administer and is ineffective against T. b. rhodesiense (<5% of total cases). 6 Combination of 2 with nifurtimox (3) (NECT) has recently led to reduced cost and workload without compromising efficacy, 7 but similar issues (adverse effects and parenteral administration) plus a lack of CNS penetration limit the two early stage drugs, pentamidine and suramin. [3][4][5] Thus, there is a compelling need for more universally effective, safe and affordable oral therapies. Two promising new agents are now in phase II/III clinical trials; fexinidazole (4) and oxaborole SCYX-7158 (5). 8,9 Nevertheless, in order to mitigate development risks and minimise the emergence of drug resistance, it remains essential to develop a pipeline of novel agents with unique mechanisms of action. 4 The nitroimidazooxazine pretomanid (PA-824, 6) has demonstrated excellent bactericidal efficacy in phase II clinical studies for tuberculosis (TB), stimulating its further appraisal in new drug combination trials. 10 Within a comprehensive backup program, in collaboration with the TB Alliance, we generated a library of more than 1000 compounds, whose assessment led to the advancement of a second generation TB candidate (TBA-354, 7) into phase I clinical evaluation. 11 We recently disclosed that phenotypic screening of some early examples against kinetoplastid diseases by the Drugs for Neglected Diseases initiative (DNDi) unexpectedly enabled the discovery of DNDI-VL-2098 (8) as a preclinical lead for visceral leishmaniasis. 12 Unfortunately, 8 exhibited poor potency against T. b. brucei (IC 50 53 mM) and it was reported 13 that 6 also had weak activity versus this parasite (IC 50 38 mM). However, unlike fexinidazole (4), 6 did not display cross-resistance to nifurtimox (3), 13 indicating that it is not activated by the same type I nitroreductase employed by 3 and 4 (implying a different mechanism of action). Therefore, as part of a wider search for improved development candidates for HAT, $900 analogues of 6 were screened by DNDi and several promising hits were unearthed. Herein, we reveal initial in vitro/in vivo profiling data on these hits, and findings from a preliminary SAR study of a nitroimidazothiazine oxide lead.
Medium-throughput screening and follow-up IC 50 testing at Scynexis 14 identified 48 active hits (IC 50 < 3 mg/mL), of which 19 were initially considered to be of potential interest (mean IC 50 < 1 mg/mL, with selectivity index > 10). Intriguingly, the most active compounds (9-12; Table 1) were either 2-nitroimidazothiazine oxides 15 or 6-nitroimidazothiazole oxides, 16 but a wide variety of other structures, including extended side chain analogues of 6, featured in this set. Since good CNS penetration is a critical requirement for the effective treatment of stage 2 HAT, 9 the 10 most potent hits were first evaluated for cell permeability in the MDCK-MDR1 assay. In this test system, apparent permeability (P app ) values !150 nm/s are indicative of high brain penetration potential provided that the transport is not affected by P-gp inhibition 17 (necessitating an absorption quotient in the range À0.1 to 0.1). Unsurprisingly, the compound with a triaryl side chain (13, MW > 500) lacked any significant permeability (P app < 0.8 nm/s), while four others (10, 14, 16, and 18) 15,18,19 gave only modest permeability values (P app < 150 nm/s) and were suggested to be P-gp substrates (absolute AQs ! 0.3). 20 On the basis of results from this training set, additional hits were selected for assessment (19)(20)(21)(22)(23) 15,16,19 and, pleasingly, all of these demonstrated a high propensity to cross the blood-brain barrier.
In order to determine the suitability of the more permeable hits for in vivo efficacy studies, we first measured their aqueous solubilities, and their tendencies to metabolise, following a 1 h incubation with CD-1 mouse liver S9 subcellular fractions. 14 Here, the most poorly soluble compounds (11, 15, 20, and 23) were also found to be the least stable, displaying half-lives of less than 70 min. Overall, the 2-nitroimidazothiazine oxides 9, 12 and 19, 15 together with the 6-amino-linked analogue of 6 (17), 21 provided the best balance of potency, stability, aqueous solubility and CNS penetration potential. This led us to probe their in vivo pharmacokinetic (PK) profiles in mice, examining concentration levels in plasma, whole blood and brain tissue following both intravenous and oral administration ( Table 2; for further experimental details, see the Supporting Information).
The most potent hit (9) exhibited an unacceptable PK profile, giving inadequate oral exposure and poor oral bioavailability (<1.5%), consistent with both its low solubility (causing unsatisfactory absorption) and more rapid metabolism. This was unsurprising, as the 4-benzyloxybenzyl analogue of 6 was known to be markedly inferior to 6 against Mycobacterium tuberculosis in vivo, despite being an order of magnitude more potent than 6 in vitro, due to similar PK issues. 22 In contrast, the 4-trifluoromethoxybenzyl congener of 9 (12) demonstrated the slowest rate of clearance of the four, and a prolonged, high exposure level above the MIC following oral dosing (Fig. 2), with good oral bioavailability (52-55%) at all three sampling sites. Moreover, the high brain:plasma concentration ratio ($3:2) presented by 12 was encouraging for CNS uptake, as required in the treatment of stage 2 HAT. 14 The sulfone derivative of 12 (19), which was produced to a significant extent in PK samples from the analysis of 12, showed reduced oral exposure, in accordance with its inferior solubility and faster rate of clearance. Given its weaker potency (5.6-fold vs 12), these results for 19 were not predictive of good in vivo activity, thus in situ oxidation of 12 should have a minimal contribution to efficacy. Finally, the most soluble hit 17 (the 6-amino analogue of 6) was notable for having the best oral bioavailability, with excellent concentration levels observed in brain tissue (2-to 3-fold higher than in plasma). However, this compound also suffered from a high rate of clearance and a rather short oral half-life (1.2-1.5 h), leading to inadequate exposure above the MIC beyond $2 h. These latter results mirrored findings from a recently reported PK-PD study of analogues of 6 against TB, in which 17 displayed a 1.3 h oral half-life in mouse lung tissue (in comparison to 4.8 h for 6), 23 effectively precluding useful in vivo activity. Hence, of the four most promising hits, only the 2-nitroimidazothiazine oxide 12 proved to be suitable for efficacy assessment in the acute infection mouse model of HAT.
One remaining matter to resolve with racemic hit 12 was which one of the four possible stereoisomers was the most active chiral form. This issue was partially clarified through a better optimised resynthesis of 12 (Scheme 1). Following side chain attachment to the racemic alcohol 42 15 (93% yield), careful oxidation of thiazine 30 with fresh m-CPBA (1.01 equiv) led to a separable mixture of 12 (75%) and a previously unidentified more polar racemic diastereomer 38 (20%) (for experimental details, see the Supporting Information).
The 1 H NMR spectra of 12 and 38 showed pronounced chemical shift differences for the H-6 resonance in particular, which was $0.4 ppm further downfield in the spectrum of 12. The sulfoxide oxygen in six-membered rings is known to exhibit an axial preference, such that the deshielding effect of the sulfoxide group on axial b-hydrogen atoms has been used to assign relative stereochemistry. 24 Hence, 12 is postulated to have the sulfoxide oxygen and H-6 in a pseudo-diaxial orientation, placing the (4-OCF 3 )ben- zyloxy side chain at C-6 in a 1,3-trans relationship to the sulfoxide oxygen. This assignment is supported by the diastereomer ratio (3.5:1) in favour of 12, which might be rationalised by an expected preference for the C-6 side chain to adopt a pseudoaxial conformation in the thiazine precursor 30 (based on the crystal structure of 6), 25 providing a steric disincentive to formation of the cis sulfoxide 38.
Preparative chiral SFC separation of the enantiomers of 12 (24 and 26) and 38 (25 and 27) facilitated the assessment of all four stereoisomers ( Table 3). The C-6 configuration of 24 and 25 was later firmly established via a known 26 chiral synthesis. The most active (6S) form 24 (IC 50 0.07 mg/mL) was 40-fold more potent than cis isomer 25, and more than 70-fold more potent than its (6R) enantiomer 26. This level of potency compared well with data reported for 5 (IC 50 0.29 mg/mL vs T. b. brucei 427) in the same Scynexis assay. 9 Compound 24 also displayed an improved selec-tivity index (>143), good aqueous solubility (106 mg/mL), and excellent metabolic stability following a 1 h exposure to human and mouse liver microsomes (respectively, 82% and 96% parent remaining).
Therefore, 24 was examined in a stage 1 HAT mouse model. 14 Briefly, dosing was orally once daily for four days, starting 24 h postinfection, and parasitemia was assessed weekly via tail vein blood smears (see the Supporting Information). Excellent activity was observed (Table 4), with 24 providing complete cures (i.e. parasite free blood smears after >30 days) to all mice at doses as low as 5 mg/kg, similar to the control drug pentamidine (given i.p. at 2 mg/kg), whereas the vehicle only mice died on day 7. The efficacy seen with 24 in this model was equivalent to the level of activity reported for 5 and $20-fold superior to the results described for fexinidazole (4), 8,9 stimulating further evaluation of this lead in a stage 2 HAT mouse model. 14 Here, oral dosing of 24 (at 12.5 to (2 mM); AQ is the absorption quotient, as defined by the equation: AQ = (P app + 918 À P app )/P app + 918. In this assay, the CNS positive drug propranolol gave P app 556 nm/s and 4 had P app 732 nm/s. 50 mg/kg once daily for seven days from day 21 postinfection) led to significant increases in survival times in comparison to untreated controls (66-70 days vs 31 days; Table 5), although cure rates were inadequate (0-20%). In contrast, 5 was 100% curative in the same CNS model at a dosage of 25 mg/kg once daily for 7 days, 9 while 4 gave an 88% cure rate in a comparable model when administered orally at 200 mg/kg once daily for 5 days. 8 Detailed in vivo studies in the benzoxaborole 6-carboxamide class have revealed that efficacy in the CNS model is heavily dependent upon the maintenance of drug concentrations in the brain for at least 15 h at levels above the MIC (defined as the lowest compound concentration that completely inhibits visible parasite growth in vitro after a 72 h incubation). 27,28 Thus, a more potent analogue of 5 without the gem-dimethyl group (SCYX-6759) required an oral dosing regimen of 50 mg/kg twice daily (b.i.d.) in order to obtain an 83% cure rate of the CNS infection, 14 due to the shorter time that its brain concentration level was at or above the MIC ($12 h vs $ 24 h for 5 at 25 mg/kg 9 ). These findings imply that a similar oral dosing regimen of 50 mg/kg b.i.d. might be required to achieve useful efficacy for 24 in the stage 2 HAT model (via more prolonged CNS exposure). Nevertheless, these initial in vivo results with 24 were still regarded as encouraging, and indicated that 2-nitro-6,7-dihydro-5H-imidazo[2,1-b][1,3]thiazine 8oxides merited further investigation as potential treatments for HAT. Specifically, as illustrated with benzoxaboroles, we considered the possibility of designing new analogues of 24 having improved potency and extended CNS exposure. On the basis of the results above and insights from previous SAR studies directed at developing a backup TB candidate to the structurally related nitroimidazooxazine 6, 19,29 we devised two preliminary strategies for optimisation of the side chain of 24: a) removal of the benzylic methylene group and b) insertion of a proximal pyridine ring (cf. 7). Notably, both strategies had the potential to improve metabolic stability, 19,29 leading to longer in vivo half-lives and better exposure levels. Furthermore, to mitigate any reduction in solubility with the first approach, we also proposed the preparation of a trifluoromethylpyridinyl ether analogue (cf . 21 and 22).
The synthetic methods employed to prepare the new nitroimidazothiazine derivatives (28,29,(32)(33)(34)(35)(36)(37)(39)(40)(41) (Table 6) were characterised by 1 H NMR, MS, melting point, and combustion analysis (or HRMS and HPLC); full synthetic procedures and characterisation data have been provided in the Supporting Information.   The new compounds and relevant comparators were screened at the University of Antwerp against a panel of four protozoan parasites (T. b. brucei, T. b. rhodesiense, T. cruzi, and L. infantum); cytotoxic effects on human lung fibroblasts (MRC-5 cells, the host for T. cruzi) were also assessed. 32 In all cases, recorded data (Table 6) are mean values derived from two or more independent experiments. For the parent thiazines (28)(29)(30)(31)(32), antitrypanosomal potency was enhanced by an order of magnitude with biaryl side chains (d and e), and this SAR pattern was maintained for the considerably less lipophilic major sulfoxide disastereoisomers (Ba-e), where 35 was the most impressive new HAT lead (T. b. brucei IC 50 0.030 mM). This lead was also highly effective against Chagas disease (T. cruzi IC 50 0.067 mM) and was the only compound to display submicromolar antileishmanial activity (L. inf IC 50 0.41 mM). In contrast, shorter linked aryl ether sulfoxides 33 and 34 were 4-to 6-fold less potent than the initial hit 12 against T. b. brucei, while their sulfone derivatives (40 and 41) were an order of magnitude inferior to sulfone 19, indicating that the original (OCH 2 ) linkage was best. In comparison to 12 (the racemic form of lead 24), racemic sulfoxide 35 displayed a 9-fold greater potency against T. b. brucei, an 11-fold higher potency against T. b. rhodesiense, and a 2.4-fold better selectivity index (MRC-5 IC 50 > 500 times larger than the HAT IC 50 ). Compound 35 also demonstrated acceptable aqueous solubility (9.9 mg/mL at pH 7 and 1260 mg/mL at pH 1), high permeability potential without P-gp mediated efflux (MDCK-MDR1 cell P app A-B/B-A 117/182 nm/s cf. P app A-B of 197 nm/s for the CNS positive drug propranolol in the same assay) and very good stability toward human and mouse liver microsomes (respectively, 78% and 72% parent remaining after 1 h). While we have not yet had the opportunity to evaluate 35 beyond this stage, these promising results certainly point to the viability of this SAR approach to provide useful new HAT leads.
In summary, this investigation set out to evaluate a nitroimidazole-based compound library related to pretomanid for possible utility against HAT. Although the hit rate was low ($2%), several compounds displayed good metabolic stability, adequate solubility, and excellent CNS penetration potential. Comprehensive mouse pharmacokinetic studies of three oxidised nitroimidazothiazines and a 6-amino-linked analogue of pretomanid identified the racemic thiazine oxide 12 as a suitable candidate for in vivo efficacy studies. The most potent stereoisomer of 12 (24) was indeed highly efficacious in the stage 1 HAT mouse model with once daily oral dosing (similar to oxaborole 5), but was less effective in a stage 2 model. While it seemed reasonable to speculate that more frequent dosing with 24 should achieve better outcomes in this latter model, we also envisaged the generation of new analogues with higher potency and longer half-lives. In preliminary SAR work, we noted that removal of the benzylic methylene was Table 6 In vitro antiparasitic activities and calculated lipophilicities of 2-nitro-6,7-dihydro-5H-imidazo [2,1- disfavoured but that adding a proximal pyridine ring (35) enhanced potency while broadly retaining other essential properties. These additional findings are very encouraging and provide a rational foundation for further development of this interesting class of antitrypanosomal agents.