Antitubercular and Antiparasitic 2-Nitroimidazopyrazinones with Improved Potency and Solubility

Following the approval of delamanid and pretomanid as new drugs to treat drug-resistant tuberculosis, there is now a renewed interest in bicyclic nitroimidazole scaffolds as a source of therapeutics against infectious diseases. We recently described a nitroimidazopyrazinone bicyclic subclass with promising antitubercular and antiparasitic activity, prompting additional efforts to generate analogs with improved solubility and enhanced potency. The key pendant aryl substituent was modified by (i) introducing polar functionality to the methylene linker, (ii) replacing the terminal phenyl group with less lipophilic heterocycles, or (iii) generating extended biaryl side chains. Improved antitubercular and antitrypanosomal activity was observed with the biaryl side chains, with most analogs achieved 2- to 175-fold higher activity than the monoaryl parent compounds, with encouraging improvements in solubility when pyridyl groups were incorporated. This study has contributed to understanding the existing structure–activity relationship (SAR) of the nitroimidazopyrazinone scaffold against a panel of disease-causing organisms to support future lead optimization.


■ INTRODUCTION
Diseases caused by protozoans and bacteria remain a major global health threat, especially in low-and medium-income countries where they affect millions of lives. Current treatment options are inadequate, attributed to issues such as toxicity, low efficacy, high cost, and rapid emergence of drug resistance. 1,2 Therefore, new and effective therapeutics are urgently needed for the antibiotic and anti-protozoan drug discovery pipelines. Nitroimidazoles were discovered in the early 1950s. 3 They have broad-spectrum activity across parasites, mycobacteria, and both Gram-positive and Gramnegative bacteria. 3 Recently, there has been a great interest in developing nitroimidazoles based on new bicyclic core scaffolds, resulting in the success of a nitroimidazooxazole, delamanid 1 and a nitroimidazooxazine, pretomanid 2 ( Figure  1) as approved drugs for treatment of tuberculosis (TB). 4,5 Both 1 and 2 are pro-drugs that share an interesting dual mode of action. Under aerobic conditions, they have inhibitory activity against the biosynthesis of mycolic acid, a key component of the mycobacterial cell wall, 6,7 while under hypoxic conditions, they release nitric oxide intracellularly, causing respiratory poisoning that kills the non-replicating bacteria. 8 Nitroimidazole 1 was found to have excellent bactericidal and sterilizing activity, with MIC values of 0.006−0.024 μg/mL against both drug-susceptible and drugresistant Mycobacterium tuberculosis under normoxic condi-tions. 6 The compound was granted conditional approval by the European Medicines Agency (EMA) in 2014 for the treatment of multidrug-resistant TB. 9 Nitroimidazole 2 was recently approved by the FDA in August 2019 as part of an all-oral combination regimen to treat drug-resistant TB, alongside bedaquiline and linezolid. 10 Drug repurposing or repositioning programs have opened a new avenue for nitroimidazoles to target other neglected diseases where R&D funding is limited. 11 Both enantiomers of 2 were shown to be active against Leishmania donovani, a causative agent of visceral leishmaniasis (VL). 12 However the R enantiomer surpassed the activity of its S counterpart in leishmania-infected macrophages and murine models. 12 A study of 1 in L. donovani revealed its superior activity against intracellular amastigotes, with >30-fold improvement of EC 50 compared to the standard drug miltefosine. 13 Another nitroimidazooxazole, DNDI-VL-2098 3 that was identified by the Drugs for Neglected Diseases initiative (DNDi), also demonstrated good efficacy in acute and chronic rodent models of VL after oral dosing. 14 However, 3 was discontinued from further studies due to safety issues. 15 Following the termination of 3, a novel 7-substituted nitroimidazooxazine, DNDI-0690 (4) was discovered within a backup program of 2 by the TB Alliance and the University of Auckland. This preclinical candidate was found to possess excellent in vitro activity against both L. donovani and Leishmania infantum as well as displaying a better safety profile than 3; hence, it was selected for further development. 16 Another nitroimidazooxazine, DNDI-8219 (5) was discovered from a library of ∼900 pretomanid analogs, with broad-spectrum activity against a range of reference strains and clinical isolates of Leishmania while displaying favorable pharmacokinetic profile. 17 Similar programs were expanded to other kinetoplastids, revealing a new thiazine oxide analog 6 that showed a complete cure in a mouse model of Trypanosoma brucei brucei infection, when dosed orally. 18 A new nitroimidazolethiazole 7 with potent activity against Trypanosoma cruzi, the causative agent of Chagas disease, was also identified, although it was not active against L. infantum. 19 This suggests the utility of testing different subclasses of bicyclic nitroimidazoles for activity in various parasite species as well as the possibility to improve their selectivity against a specific organism.
The market approvals of nitroimidazoles are encouraging, but there is still room for improvement. Delamanid 1 suffers from poor water solubility and low absorption, requiring twicedaily dosing for the treatment of multidrug-resistant TB. 20 Pretomanid 2 has better PK properties with daily dosing, but it is less potent compared to 1. A second-generation nitroimidazooxazine, TBA-354 8, was reported to have superior antitubercular activity than 2 and better metabolic stability than 1, 21 but it was discontinued from phase I clinical trials due to neurotoxicity. 22 To address some of these drawbacks, various strategies have been applied, mostly focusing on the modification of the aryl side chain of the imidazo-oxazole/ oxazine ring. 23−25 A library of new analogs of 2 with extended side chains and different biaryl linkers was synthesized by Palmer et al. to enhance the solubility and oral activity in chronic M. tuberculosis infection. 23 Most of the lipophilic and polar linkers showed comparable or improved in vitro antitubercular activity, whereas the introduction of hydrophilic groups and replacement of phenyl with pyridine rings were beneficial in improving the aqueous solubility. 23 Similar measures were also employed by Yempalla et al. 24 and Thompson et al. 25 to enhance the physicochemical/pharmacological properties of nitroimidazooxazole against M. tuberculosis and L. infantum, respectively.
Our previous work identified a new bicyclic nitroimidazole scaffold with promising antitubercular and antiparasitic activity, namely, nitroimidazopyrazinone 9. 26 This scaffold was derived from the 4(5)-nitroimidazoles that were previously developed, 27 taking inspiration from both metronidazole and 2. The O-alkylated nitroimidazopyrazine regioisomers of 9 were also isolated during the reaction, but they were inactive against M. tuberculosis, even though they retained antiparasitic activity against Giardia lamblia and T. b. brucei. 26 This demonstrated the significance of the bicyclic core in determining their selectivity profile toward different organisms. Similar to other bicyclic nitroimidazoles such as 1 and 2, most of the potent nitroimidazopyrazinones displayed poor aqueous solubility, with a few exceptions for some of the actives against T. b. brucei. 26 This led us to the current study, in which we have now designed a new series of analogs in an attempt to improve solubility without compromising the activity against a panel of disease-causing organisms. We investigated three approaches, all focused on modification of the substituent at R 1 without Journal of Medicinal Chemistry pubs.acs.org/jmc Article altering the bicyclic imidazopyrazinone core ( Figure 2). Modified compounds were tested for antibacterial, antiprotozoal, and antifungal activity in an effort to identify hit compounds against these organisms. We chose not to impart any substitutions at R 2 and R 3 as these modifications were previously shown to be less favorable for antitubercular activity, despite showing tolerable potency against T. b. brucei. 26 In the first approach, a small set of analogs was prepared where the linker was modified by introducing ketone, amide, and ether functionalities with varying lengths to the benzylic methylene moiety. In the second approach, less lipophilic heterocycles were used to replace the phenyl group. Lastly, extended side chains such as biphenyl analogs and their bioisosteres were synthesized, with and without the incorporation of less lipophilic heterocycles.

■ CHEMISTRY
From our previous SAR findings, aromatic substituents at R 1 were preferred over polar side chains and bulky aliphatic groups. 26 Therefore in series 1, we sought an alternative approach to reduce the lipophilicity by varying the linker component without altering the active aromatic moiety. The key nitroimidazopyrazinone intermediate 16 was first synthesized in four steps as reported, 26 starting from the commercially available imidazo-carboxylic acid 12 (Scheme 1). These involved nitration using concentrated H 2 SO 4 and HNO 3 , formation of nitroimidazole carboxamide 14, alkylation with bromoacetaldehyde diethyl acetal under basic conditions, and finally cyclization between 1′ imidazole and 2′ free amide nitrogen to afford the bicyclic core 16. 26 Different linker variants (17a−g) were synthesized from 16 through alkylation with alkyl halides using basic carbonate (Cs 2 CO 3 or K 2 CO 3 ) as a catalyst. Ketone-linked products (17a−c) were synthesized from bromoacetophenone derivatives at room temperature, though with poor yield (23−41%). For amide linker variants (17d−f), intermediate bromo-N-substituted acetamides (11ac) were first synthesized via acylation of various amines and 2bromoacetyl bromide in the presence of trimethylamine 28 followed by N-alkylation under microwave irradiation. Hydroxyl derivative (17h) was prepared from 17g by deacetylation using K 2 CO 3 in methanol, which was then (iii) oxalyl chloride, catalytic DMF, DCM, 0°C → rt, then concentrated NH 4 OH, 0°C → rt; (iv) bromoacetaldehyde diethyl acetal, K 2 CO 3 , μW 180°C; (v) 2 M HCl/1,4-dioxane, μW 120°C; (vi) various alkyl bromides, Cs 2 CO 3 or K 2 CO 3 , DMF, μW 80−100°C; (vii) K 2 CO 3 , MeOH, rt; (viii) various benzyl bromides, NaH, DMF, 0°C → rt. a Reported in a previous study. 26 coupled with the desired benzylic bromides in the presence of NaH to form the ether-linked analogs (17i−k). For series 2, less lipophilic aromatic rings with hydrogen bond donors/acceptors, such as pyridine, thiazole, and imidazole were introduced. Two different routes, as described in Scheme 2, were utilized to synthesize 26a−k. In the first route, the chloride intermediates 18−20 were converted from the alcohol derivatives in thionyl chloride, 29 whereas 21 was formed by bromination using phosphorus tribromide. 30 These halide intermediates were used immediately in the following step due to their instability. Benzimidazoles 24 and 25 were made from 22 and 23 and chloroacetic acid by the Philip method, 31 which were easily isolated in high purity by vacuum filtration after neutralizing with NaHCO 3 . Subsequently, Nalkylation was performed between 16 and the halide intermediates to afford the final products (26a, c−f, h) in 42−75% yield after purification. Alternatively, 26b, g, i−k were prepared via route 2, which involved the formation of 4(5)nitroimidazole carboxamides 27b, g, i−k of different derivatives. 27 Alkylation of 27b, g, i−k with bromoacetaldehyde diethyl acetal under microwave conditions provided 28b, g, i−k followed by cyclization in acid to give the bicyclic products in 13−83% yield.
For the third approach, a variety of biaryl side chain intermediates (33b−m) were synthesized to produce the series 3 analogs 34b−m, while biphenyl 34a was prepared from commercial source 33a (Scheme 3). Aldehydes 29a−c were first reacted with boronic acids 30a−e via palladium-catalyzed Suzuki coupling to form 31b−m, 32 which were then purified by silica chromatography. Reduction of 31b−m using sodium borohydride produced alcohols 32b−m under mild conditions (room temperature) and good yield. Chlorination under treatment of thionyl chloride furnished 33b−m, which were used directly in the final step of alkylation. Synthesis of 34n,o was also accomplished using similar chemistry, starting from the reduction of aldehydes 31n,o. While the isolation yield of these extended biaryl analogs were varied (37−98%), they were synthesized rapidly (less than 1 h) under microwave conditions and less impurities were observed by LC-MS. Compound 34p was synthesized from 4(5)-nitroimidazole carboxamide 36 through acylation with (4-(piperidin-1-yl)-phenyl)methanamine 35 followed by N-alkylation and cyclization under a similar reaction condition as 27b, g, i−k.
histolytica as their first generation and the monocyclic intermediates, nitroimidazole carboxamides, 27 were active against these intestinal parasites. Therefore, selected compounds from each series were screened for activity against these protozoans. Given other nitroimidazoles (e.g., metronidazole) have antibacterial activity beyond M. tuberculosis, we assayed their antibacterial properties against representative ESKAPE pathogens (Escherichia coli, Staphylococcus aureus, Klebsiella pneumoniae, Pseudomonas aeruginosa, and Actinetobacter baumannii) as well as their antifungal potential against Cryptoccoccus neoformans and Candida albicans were assessed. Mammalian cell viability was evaluated against human embryonic kidney (HEK293) cells in order to establish a preliminary selectivity index for antimicrobial activity versus mammalian toxicity. For compounds tested against T. cruzi, cytotoxicity against 3T3 host cells was also determined. Lipophilicities of these analogs were estimated using Pipeline Pilot (Accelrys, version 9.1.0.13). Solubility was measured in water and phosphate-buffered saline (PBS), pH 7.4 using HPLC with UV analysis. Representative compounds with weak base functionalities were also tested for their aqueous solubility in acetate buffer, pH 4.5 and 0.1 M HCl, pH 1. The low pH mimics the gastric fluid environment, which is beneficial to enhance dissolution of weakly basic drugs. 35 Series 1 (Table 1). For the first series 17a−k, insertion of ketone, amide, and ether groups to the methylene chain led to overall reduced lipophilicity based on Pipeline Pilot AlogP calculations ( Table 1). The presence of oxygen in ether and ketone linked molecules functions as a strong hydrogen bond acceptor, whereas the amide linker serves as a strong hydrogen bond acceptor and as a modest hydrogen bond donor. Compared to our reported analog 16a with a MIC of 0.5−1 μg/mL against H37Rv strains under aerobic growth, 17a, 17d, 17e, and 17i were inactive (MIC normoxia ≥ 32 μg/mL), although 17i showed ∼11-fold of improvement in its solubility. Some activity was restored with the addition of an electrondonating 4-methyl group at the phenyl ring. However, this more polar series generally showed unfavorable antimycobacterial activity ( Table 1). As many of the recently developed anti-TB agents are highly lipophilic, this suggests the importance of lipophilicity in a molecule to penetrate into the hydrophobic phase of mycobacteria cell wall. 36 Determination of antitrypanosomal activity disclosed a different SAR. Amide linkers were overall well tolerated with respect to T. b. brucei activity, giving at least an 8-fold improvement of activity compared to the methylene linker (incomplete inhibition at 16 and 40 μM). Compound 17d (−CH 2 CONH−) with an amide bond directly linked to the phenyl side chain (IC 50 = 0.86 μM) was ∼3-fold more potent than 17e (−CH 2 CONHCH 2 −) with an extra carbon length (IC 50 = 2.4 μM). Addition of a 3-trifluoromethoxy group at the phenyl ring was also favorable (∼7-fold, IC 50 = 0.36 μM). Though the presence of amide linker lowered the calculated lipophilicity (ΔAlogP ≈ −0.94 units), there was little or no improvement in their aqueous solubility. The introduction of a more soluble ether group linker displayed only a mild effect on T. b. brucei activity compared to the parent analogs (16a−c). These ether-linked derivatives 17i−k were the only actives in this series against T. cruzi, with IC 50 values ranging from 1.3 to 2.3 μM, comparable activity to one of the standard drugs nifurtimox (IC 50 = 1.4 μM) and more potent than benznidazole (IC 50 = 5.2 μM). Though they were less active than the parent analog 16b, these series demonstrated improved solubility in some cases and had good selectivity indices of >43 to >77.  (Table 2). Overall, compounds with less lipophilic heterocycles replacing the aryl substituent displayed significant improvement of solubility compared to the phenyl substituent. However, in most cases, their antimycobacterial activity was compromised. For example, compound 26a containing a thiazol-5-yl substituent was >50-fold more soluble than 16a, though its activity against M. tuberculosis under normoxic conditions was reduced by 32-fold ( Table 2). Replacement of the phenyl ring with a more soluble pyridine group was also detrimental to its activity. However, addition of a 4-CF 3 electron-withdrawing group managed to restore some of the activity (MIC normoxia = 2 μg/mL cf. MIC normoxia of 16a = 0.5−1 μg/mL) and concurrently provided superior aqueous solubility (30 μM cf. 16a: 3.1 μM). It was found that compounds with negative AlogP values generally possessed poor MIC normoxia ≥ 16 μg/mL. Analogs with positive AlogP in this series remained active (MIC normoxia = 2−4 μg/mL), with the exception of compound 26f that displayed solubility problems even in the DMSO stock solution. None of the active analogs surpassed the activity of the parent compounds (for example, 16a−d), suggesting that an alternative strategy is required to achieve a desirable compromise between solubility and bioactivity.
We next investigated the SAR of this series against T. b. brucei. As in series 1, in most cases, the less lipophilic heterocycles were well tolerated, with an IC 50 of <10 μM. Compound 26j with a quinolone substituent was the most active in this series (IC 50 = 0.89 μM) and gave a substantial enhancement of solubility (>8-fold) compared to 16a. The weak basic nature of 26j allowed for further improvement of solubility under acidic conditions (70 μM at pH 4.5 and > 200 μM at pH 1). It is noted that 26j was also active against T. cruzi at submicromolar concentrations (IC 50 = 0.54 μM), and selectivity index of >136 to 3T3 host cells (Supporting Information, Table S2). As this analog was moderately active against M. tuberculosis (MIC normoxia = 4 μg/mL), this indicates the possibility to identify compounds with broad-spectrum activity and low toxicity against mammalian cells that could potentially treat both tuberculosis and kinetoplastid infections.
Series 3 (Table 3). Following the precedent set by several new bicyclic nitroimidazoles with biaryl side chains such as 4 and 8 that are in pre-clinical development against leishmaniasis and tuberculosis, respectively, 16,21 a similar strategy was employed in the third series. As suggested from the SAR for series 1 and 2, compounds with less lipophilicity had less favorable antimycobacterial activity. Therefore, extended and highly lipophilic biphenyl side chains were explored to understand whether this property is the main driver. It was also postulated that the extended side chains could interact more efficiently with the hydrophobic pocket in Ddn, the nitroreductase enzyme responsible for the bioactivation of bicyclic nitroimidazoles 1 and 2. 37 To counter the solubility issue, bioisosteres were also introduced to replace one of the phenyl rings in this series. The effect of different biaryl geometries with 4′ and 3′-linked biaryls as well as the position of the substituents were investigated. Based on the results of the previously developed monoaryl series, we included the  Table  S2. g For compounds that did not completely inhibit growth at the concentration tested, the percentage inhibition (% I) is given. ND, not determined. Generally, the biphenyl analogs 34a−d from Table 3 show that these conformationally rigid extended side chains did not contribute to a better activity against M. tuberculosis. In fact, they were 16-to 64-fold less active than the monoaryl analogs 16a−d. This observation is different from what has been reported for the biphenyl structures of analogs of pretomanid 2, in which the para-linked compounds retained or improved their antitubercular activity compared to the monoaryl 2. 38 This suggests a different binding mode of the nitroimidazopyrazinone scaffold 9 in the active site compared to 2, which is not unexpected given the planar nature of 9 compared to 2. There was a preference for substitution at the meta position over para at the terminal ring of the 4′ biphenyl analogs. A similar trend was observed for the less lipophilic phenylpyridine analogs (ΔAlogP ≈ −0.7 unit), where the 3-trifluoromethoxy congener 34i was 8-fold more active than its 4-trifluoromethoxy counterpart. Overall, the phenylpyridines 34f−i gave similar or 2-fold reduced activity than the 4′-biphenyl analogs 34a−d, though the solubility (at pH 1) was improved for some of these derivatives. The position of the nitrogen atom at the pyridine ring was also found to be important for its bioactivity. Replacement of the 3-pyridine ring in 34g by 2-pyridine (34j) substantially improved its antimycobacterial activity, from an MIC of 32 μg/mL to 0.0625−0.125 μg/mL against the replicating M. tuberculosis. However when M. tuberculosis was grown under hypoxic conditions (0.1% oxygen), 34j exhibited only marginal inhibitory activity, though still better than 34g. Under acidic conditions (pH 1), 34j possessed >3-fold better solubility than 34g.
According to the general solubility equation developed by Jain and Yalkowsky, a compound can exhibit poor solubility due to high lipophilicity and high stability in the crystalline state. It is estimated that altering log P by 1 unit or melting point by 100°C will change solubility by 10-fold. 39 Hence, an alternative strategy was attempted by disrupting their molecular symmetry. The crystal packing of a molecule is influenced by its molecular symmetry and planarity, and improved solubility can be achieved by decreasing the efficiency of its crystal packing to lower the melting point. 40 In the 3′-linked phenylpyridines 34k−m, we have observed at least 3-to >16-fold of improvement in solubility at pH 1 compared to 4′-linked phenylpyridines, with the 3′-substitution disrupting the potential for planar packing. Compounds 34k−m also showed 2-to 128-fold better antitubercular activity compared to their 4′-linked analogs. Among them, 34l displayed an MIC of 0.0625−0.125 μg/mL, which was equipotent with the parent analog 16c. As opposed to the 4′-linked biaryls, the para-substitution (34k) was preferred over its meta group (34m) in terms of potency and solubility.
Swapping the positions of the less lipophilic ring in the biaryl system led to substantial effects on potency and physicochemical properties ( Figure 3). Biaryl compound 34e with a terminal 3-pyridine ring displayed 64-fold of improvement in antitubercular activity under aerobic conditions as compared to 34f with a terminal phenyl ring. Compound 34e was also more soluble than 34f in both water and PBS, pH 7.4, and its solubility was further improved at pH 4.5 (19 μM) and pH 1 (>200 μM). When pyridine was replaced by a thiazole group at the terminal ring (34n), there was a complete loss of activity, indicating that the antitubercular activity was strongly dependent on the nature of the heterocycles. In spite of having similar AlogP values, 34n was also at least 3-fold less soluble than 34e in water and PBS. When the position between thiazole and phenyl rings was inverted (34o), the antitubercular activity improved at least 32-fold, reaching an MIC normoxia of 0.5−1 μg/mL. Removing the aromaticity by replacing the pyridine ring with a piperidine ring (34p) resulted in 2-to 4-fold reduction of activity against replicating In comparison to the monoaryl compounds, the biaryl series demonstrated superior activity against both T. b. brucei and T. cruzi. Of a total of 16 biaryl analogs synthesized, 14 were identified as active hits, with mean IC 50 values of <1 μM and selectivity indices of >10 for parasitic activity compared to mammalian cell cytotoxicity. More than half of these hits were able to achieve IC 50 at ≤0.1 μM. Contrary to the SAR in mycobacteria, 4′-phenylpyridines 34f−i possessed better activity against T. b. brucei than the 4′-biphenyl side chains 34a−d. Compounds 34g-i were also ∼1.1to 8.4-fold more active than 3′-linked phenylpyridines 34k−m, with an IC 50 range of 0.008−0.026 μM. A similar SAR trend was observed in T. cruzi, though their activity was less compared to T. b. brucei in most of the cases. Compound 34j with 2-pyridine ring was identified as the most potent candidate, giving an impressive IC 50 of 0.002 μM and 0.016 μM against T. b.
brucei and T. cruzi, respectively, with 100% activity at E max concentrations (maximal % inhibition plateau generated in dose−response curves). This compound also exhibited selectivity indices against HEK293 cells (20 h assay) of >37,000 and >4625, and selectivity of >2288 to 3T3 host cells (69 h assay, Supporting Information, Table S2). It should be noted that 34j was one of the most active antitubercular hits, suggesting a broad-spectrum activity. Altering the position of the thiazole ring from 34n to 34o enhanced the activity against both T. b. brucei and T. cruzi, whereas a marginal effect was observed when swapping pyridine from 34e to 34f. When the terminal 3-pyridine ring was replaced by a saturated heterocycle (piperidine), the antitrypanosomal activity of 34p was diminished.
to the trends shown in M. tuberculosis, modification of the methylene chain resulted in drastic loss of activity (Table 4). Replacing the benzyl substituent with less lipophilic monoheterocycles (Series 2) also proved to be unfavorable. Fused rings such as quinolone (26j) and benzodioxole (26k) managed to restore some of the activity, with IC 50 values of 7.1 and 2.4 μM, respectively. Most of the tested biaryl analogs remained active in the primary screen at 50 μM. However, some inconsistent SAR results were found when determining their IC 50 values. The biphenyl compound 34a achieved 10fold better potency than its monophenyl counterpart (IC 50 = 0.5 μM cf. IC 50 of 16a = 5 μM), whereas addition of a 3-OCF 3 substituent (34d) reduced activity by >20-fold. Follow-up study is hence required in the future to determine the inhibitory concentration of more analogs from this series.
Besides antigiardial activity, selected analogs from each series were also screened against E. histolytica, ESKAPE bacteria strains and two fungal pathogens. Similar to our first generation of nitroimidazopyrazinones, none of the tested compounds were active against these organisms. We have also counter screened these compounds against mammalian cell line HEK293 and minimal reduction in cell viability was seen up to the highest tested concentrations (>50 or >100 μM, respectively, depending on the solubility in DMSO stock, Tables 1−3).
Plasma Protein Binding and Microsomal and Plasma Stability. To determine the drug-like properties of these new series, a small subset of compounds was evaluated in vitro to determine their plasma protein binding (PPB), microsomal stability, and plasma stability ( Table 5). The more soluble analogs 17k and 26j demonstrated a reduced propensity to bind to plasma protein (72−79% PPB in human and mouse plasma) compared to other compounds. The biaryl series were generally highly bound to plasma, especially compounds 34g, 34j, and 34k that demonstrated >99% PPB due to their high lipophilicity (AlogP > 3). On the other hand, phenylpyridine 34e and phenylpiperidine 34p, with better solubility and lower lipophilicity, demonstrated a slightly lower PPB (93−99%). A good correlation between the compound lipophilicity and PPB was identified, suggesting the potential to utilize this parameter to modulate protein binding. However, it is also known that free drug concentration is not merely determined by PPB but also by hepatic intrinsic clearance after oral dosing. 41 Thus, we investigated the metabolic stability of these analogs in human liver microsomes (HLM). The biaryl compounds tested were highly stable up to 2 h. Replacing the terminal aryl ring with piperidine (34p) reduced the microsomal stability from >95% to 37%. Similar to some of the reported pretomanid analogs, 42 benzyl ether substituent (17k) was highly susceptible to metabolism, with 8.5% of intact compound remaining at 2 h. We have also expanded the metabolic assay to CD-1 mouse liver microsomes (MLM) for two hit compounds, 34e and 34j, prior to assessment in future in vivo studies against T. brucei and T. cruzi. The most potent compound 34j displayed excellent stability in both HLM and MLM. However, interspecies variation was observed for phenylpyridine 34e, with 20% of intact compound remaining at 2 h in MLM compared to 96% in HLM. Similar to the first generation of nitroimidazopyrazinones, 26 these new series remained highly stable in plasma (>90% at 2 h) regardless of their side chains.

■ CONCLUSIONS
Current therapeutics against neglected tropical diseases are imperfect due to toxicity, inadequate efficacies against all subspecies/species, and prolonged regimens with poor patient compliance. M. tuberculosis has developed resistance rapidly, leading to the emergence of multidrug-and extensively drugresistant TB with low cure rates. To improve treatment regimens in an effort to reduce poor patient compliance, low dosing frequency, reduced toxicity, and easy administration are required. Oral, affordable drugs are preferred as they are more convenient to be used, thus increasing compliance especially in developing countries where treatment is often far from the patient. However, many new chemical entities are poorly soluble, which makes them difficult to achieve desired therapeutic concentration after oral dosing.
In this study, we aimed to optimize the solubility and activity profiles of a new bicyclic nitroimidazole subclass, namely, nitroimidazopyrazinone, as an oral treatment against tuberculosis and parasitic infections. Three new exploratory series were synthesized, featuring a range of linker groups, heterocycles, and extended biaryl side chains. Modification of the methylene chain of 16 with polar linkers was not favorable for activity against M. tuberculosis, though some of them were active against the trypanosome species tested. Similarly, attempts to replace the phenyl substituent with less lipophilic heterocycles resulted in significant improvements of solubility but with compromised antitubercular activity. The series with biaryl extended side chains provided the most promising outcome in optimizing both the potency and solubility while retaining good selectivity index of >10 to mammalian cell lines. Several phenylpyridine analogs showed improved solubility at low pH conditions without compromising their activity against M. tuberculosis. It was also found that the meta-linked analogs were superior to the more linear, para-linked counterparts, in terms of their solubility and antimycobacterial activity. The biaryl compounds were highly active against both T. b. brucei and T. cruzi, with an increase in activity by 1−4 orders of magnitude in comparison to the monoaryl series. Compound 34e with a terminal pyridine ring combined both better solubility profile (>200 μM at pH 1) with good bioactivity against M. tuberculosis and trypanosomes but displayed a shorter half-life in mouse microsomes. The 2-pyridine 34j was identified as the most potent hit, demonstrating submicromolar to nanomolar inhibitory concentrations against M. tuberculosis, T. b. brucei, and T. cruzi while showing low cytotoxicity against HEK293 cells (CC50 > 74 μM, or selectivity indices of >510 to >37,000). Compound 34j also possessed good metabolic stability in both microsomes and plasma, despite having high PPB levels. As many clinical drugs are highly bound to plasma, it is not recommended to optimize this parameter in early-stage drug discovery.
This study highlights the potential of nitroimidazopyrazinone as a promising bicyclic subclass to be developed as antiinfective agents. It was found that some of the potent antitubercular biaryl compounds were highly active against both T. b. brucei and T. cruzi, unveiling their broad-spectrum activity for treating multiple infections including neglected tropical diseases. Several compounds were also moderately active against G. lamblia, albeit with no activity against E. histolytica. Similar to the activity profile of pretomanid 2, none of these nitroimidazopyrazinones was active against ESKAPE and fungal pathogens (MIC > 32 μg/mL). For further Journal of Medicinal Chemistry pubs.acs.org/jmc Article development of this series, it appears that both broad-spectrum (antitubercular and antiparasitic) and narrow-spectrum strategies are possible, but it is less clear which approach is most desirable. Co-infection with both mycobacteria and trypanosomes happens in some regions of the world, where a broadspectrum treatment could be useful. However, this is offset by the possibility that widespread use against one organism could inadvertently lead to the development of resistance against the other. Since nitroimidazoles are pro-drugs that require bioactivation of their nitro group to exert the activity, resistance can arise when there is a change of activity or mutations of the reductive enzymes. 3 In some cases, this resistance can be overcome by other nitroimidazoles with different mode of action. Further studies are needed to assess the cross-resistance profile of the nitroimidazopyrazinones. In vitro ADME and cytotoxicity studies have revealed favorable drug-like profiles for the hit compounds, which warrant advancement to in vivo testing to determine pharmacokinetic profiles and assess activity in both acute and chronic infection models of M. tuberculosis, T. b. brucei, and T. cruzi.

■ GENERAL EXPERIMENTAL SECTION
All the reagents and solvents were obtained from commercial sources and were used without further purification. Progression of reaction was monitored by TLC using Merck alumina sheets pre-coated with silica gel 60 F254 and was visualized using a UV lamp. Reactions requiring anhydrous conditions were performed under an inert atmosphere of nitrogen. Purification of compounds was done using a Biotage Isolera or Grace Reveleris chromatography system. NMR data were collected at 298 K on a Varian Unity 400 MHz or Bruker Advance 600 MHz spectrometer. Chemical shifts were measured relative to tetramethylsilane or the residual solvent signals (DMSOd 6 ) as the internal reference. Data are presented as follows: chemical shift (ppm), multiplicity (s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet, bs = broad singlet) and coupling constant (Hz). All the peaks for final products were assigned based on onedimensional and two-dimensional NMR analysis. Analytical LC-MS was performed on a Shimadzu LCMS-2020 using 0.05% formic acid in water (solvent A) and 0.05% formic acid in acetonitrile (solvent B) as the mobile phase. General Procedure for Alkylation of Imidazopyrazinone. Method B: To a stirring suspension of imidazopyrazinone (1 equiv) in dry DMF (20 vol) was added 3 equiv of cesium carbonate or potassium carbonate, followed by addition of 1.2−1.5 equiv of benzyl bromide in dropwise. The reaction was either stirred at room temperature or microwaved at 80−100°C.
Method C: To a stirring suspension of imidazopyrazinone (1 equiv) in dry DMF (20 vol) at 0°C was added 1.5 equiv of 60% sodium hydride followed by addition of 1.2−1.5 equiv of benzyl bromide dropwise. The reaction was stirred at room temperature overnight under N 2 .
General Procedure for Chlorination of Alcohol. Method D: Alcohol (1 equiv) was added with DCM (20 vol) and was chilled at 0°C . Thionyl chloride (2−3 equiv) was added dropwise, and the reaction was continued to stir at room temperature for 1−24 h. The mixture was then diluted with NaHCO 3 (200 vol) and extracted with EtOAc (3x). The organic layer was collected and dried over MgSO 4 before removing the volatiles in vacuo.
General Procedure for Amide Coupling. Method E: 4-Nitro-1Himidazole-2-carboxylic acid (1 equiv) was added with anhydrous DCM (20 vol) and was chilled at 0°C under N 2 . Oxalyl chloride (2 equiv) was added dropwise over 10 min followed by addition of catalytic DMF. The reaction was allowed to warm to room temperature and was continued to stir overnight. Additional oxalyl chloride and DMF was added to the reaction mixture when necessary. Volatiles were removed in vacuo, and the excess oxalyl chloride was removed by co-evaporation with toluene two times. The resulting acid chloride crude solid (1 equiv) was then added immediately to anhydrous DCM (15 vol) and triethylamine (2 equiv) in an ice bath. The blue-purple solution was then added dropwise to the respective amine (1.2 equiv) in 5 vol of anhydrous DCM at 0°C under N 2 .
General Procedure for Alkylation of Imidazo Carboxamide. Method F: A mixture of imidazo carboxamide (1 equiv) and potassium carbonate (3 equiv) in dry DMF (20 vol) was reacted with bromoacetaldehyde diethyl acetal (1.5 equiv x 2) in a microwave reactor at 150°C (15 min x 2). If necessary, the reaction was heated for another 15 min at 150°C to consume the starting material.
General Procedure for Ring Closure to Form Imidazopyrazinones. Method G: To a stirring solution of imidazo carboxamide (1 equiv) in 1,4-dioxane (10 vol) was added 2 M aq. HCl (10 vol). The reaction was reacted at 120°C under microwave for 20 min.
General Procedure for Suzuki Coupling Reaction. Method H: To a stirring mixture of 4-bromobenzaldehyde or bromonicotinaldehyde (1 equiv) in THF was added 3 equiv of potassium carbonate in distilled water followed by addition of 0.08 equiv of tetrakis-(triphenylphosphine)palladium (0). The mixture was stirred at room temperature for 5 min. Boronic acid (1.1 equiv) was then added, and the reaction was heated at 80°C for 18 h.
General Procedure for Reduction of Aldehyde to Alcohol. Method I: Sodium borohydride (1.3 equiv) was added in one portion to a stirring solution of aldehyde (1 equiv) in 20 vol of ethanol that was chilled to 0°C. The mixture was continued to stir at room temperature for 30 min before the solvent was removed in vacuo. The crude product was diluted with 100 vol of distilled water and extracted with chloroform (3x). The organic layer was collected, washed with brine, and dried with MgSO 4 before removing the volatiles in vacuo.  278 mmol, 1 equiv) was reacted with phenacyl bromide (82.9 mg, 0.416 mmol, 1.5 equiv) at room temperature for 30 min according to the general procedure method B. After the reaction was completed as analyzed by LC-MS, the mixture was diluted with distilled water (15 mL) and the precipitate was collected by filtration. The precipitate was then purified by recrystallization using hot ACN/MeOH to yield the desired product, 2-nitro-7-
Plasma Protein Binding. Plasma protein binding (PPB) was performed using an ultrafiltration method. 58,59 Both human and mouse plasma were obtained commercially from BioReclamationIVT. Test compounds (5 μM) were incubated in 100% human plasma at 37°C for 30 min. For unfiltered samples, an aliquot (50 μL) was removed, diluted with PBS (50 μL), and quenched with ice-cold precipitating solution comprising a 0.5 μM carbutamide MS internal standard in acetonitrile:methanol:formic acid (1:1:0.001). Samples were incubated at 4°C for 30 min, then centrifuged at 14,000 × g for 8 min before the clear supernatant was transferred to a vial for LC− MS/MS analysis. For filtered samples, the plasma sample (250 μL) was filtered using Amicon Ultra-0.5 Centrifugal Filter Devices 30K NMWL (Merck Millipore) at 14,000g for 7 min, and then an aliquot (50 μL) was processed as described for unfiltered samples. The fraction of unbound compound was calculated by determining the concentration of the filtered sample and the concentration of unfiltered sample. All samples were tested in triplicate with sulfamethoxazole as a control.
Solubility Determination. Stock compound solution (20 mM in DMSO) was aliquoted into water, phosphate-buffered saline (PBS), pH 7.4 and 0.1 M HCl (pH 1), respectively, to a final concentration of 200 μM, 1% DMSO. After 24 h of incubation in a shaking incubator at room temperature, 130 rpm, samples were filtered using centrifuge filter tubes (Corning Costar Spin-X centrifuge tube filters, CLS8169) at 8000 rpm for 1 min. The filtrates were further diluted with acetonitrile (1:1, v/v) prior to analysis using LC-UV as detailed in the General Experimental Section. The solubility was determined based on the peak area at a UV absorbance of 254 nm, with reference to the standard calibration curve prepared from 20 mM DMSO stock. Compounds and standards (caffeine and pretomanid) were prepared in duplicate, and each sample was analyzed in duplicate by LC-UV.
LC-MS/MS detection and analysis parameters for plasma protein binding, microsomal and plasma stability; (Table S1) mass spectrometer parameters of analogues tested for their plasma protein binding, microsomal and plasma stability; and 1 H and 13 C NMR spectra (PDF) Molecular formula strings (XLSX) ■ AUTHOR INFORMATION ESKAPE bacteria and fungal pathogens. The antimicrobial screening performed by CO-ADD was funded by the Wellcome Trust (UK; Strategic Funding Award: 104797/Z/ 14/Z) and The University of Queensland (Australia; Strategic Funding Award). We thank Emily Kennedy for assisting with the T. cruzi in vitro assays and Elouise Gaylord for assisting with the T. b. brucei in vitro assays. We also thank Angelo Frei for his help in graphic design. C.W.A. was supported by an Australian Government Research Training Program scholarship. M.L.S. was the recipient of a Griffith University Postdoctoral Fellowship (GUPF) award. A.D. was supported by the grants 1KL2TR001444, R21AI141210, R21AI133394, and R21AI146460 from the NIH. M.A.B. was supported in part by Wellcome Trust Strategic Award 104797/Z/14/Z. M.A.C. is a NHMRC Principal Research Fellow (APP1059354) and holds a fractional professorial research fellow appointment at The University of Queensland, with his remaining time as the CEO of Inflazome Ltd., a company developing drugs to address clinical unmet needs in inflammatory disease. Cell lines (bacteria, fungi, and mammalian) were sourced from the American Type Culture Collection (ATCC).