Pentacyclic Nitrofurans with In Vivo Efficacy and Activity against Nonreplicating Mycobacterium tuberculosis

The reductively activated nitroaromatic class of antimicrobials, which include nitroimidazole and the more metabolically labile nitrofuran antitubercular agents, have demonstrated some potential for development as therapeutics against dormant TB bacilli. In previous studies, the pharmacokinetic properties of nitrofuranyl isoxazolines were improved by incorporation of the outer ring elements of the antitubercular nitroimidazole OPC-67683. This successfully increased stability of the resulting pentacyclic nitrofuran lead compound Lee1106 (referred to herein as 9a). In the current study, we report the synthesis and antimicrobial properties of 9a and panel of 9a analogs, which were developed to increase oral bioavailability. These hybrid nitrofurans remained potent inhibitors of Mycobacterium tuberculosis with favorable selectivity indices (>150) and a narrow spectrum of activity. In vivo, the pentacyclic nitrofuran compounds showed long half-lives and high volumes of distribution. Based on pharmacokinetic testing and lack of toxicity in vivo, 9a remained the series lead. 9a exerted a lengthy post antibiotic effect and was highly active against nonreplicating M. tuberculosis grown under hypoxia. 9a showed a low potential for cross resistance to current antitubercular agents, and a mechanism of activation distinct from pre-clinical tuberculosis candidates PA-824 and OPC-67683. Together these studies show that 9a is a nanomolar inhibitor of actively growing as well as nonreplicating M. tuberculosis.


Introduction
Mycobacterium tuberculosis remains an important global pathogen that is believed to have infected a third of the world's population and kills over a million persons annually [1]. The US Centers for Disease Control and Prevention recommended regime for drug sensitive tuberculosis (TB) infections consists of rifampin, isoniazid, pyrazinamide and ethambutol for an initial two months followed by continued four month treatment with isoniazid and rifampin. Poor patient compliance to this lengthy drug regimen can lead to patient relapse and promotes the emergence of drug-resistant strains, which have become a major treatment challenge and global health threat. Tuberculosis strains resistant to isoniazid and rifampin, two of the most powerful antituberculosis agents, are classified as multi drug-resistant (MDR TB). Extensively drugresistant (XDR) strains, which are additionally resistant to a fluoroquinolone, and an injectable second-line drug, have been reported in over 80 countries [2]. MDR/XDR TB dramatically increases the medical burden, requiring extended treatment lengths with drugs that are less effective and show significant toxicity [3,4]. It has been widely postulated that the key to decreasing treatment times and the spread of drug-resistant tuberculosis is the development of new therapeutics that target dormant bacilli, which are metabolically inactive cells that display phenotypic (non-genetic) resistance to most drugs [5]. Dormant bacteria are believed to reside within the nutrient and oxygen limited environment of necrotic lesions. Therefore, it is imperative that novel antitubercular agents target these persisting bacterial subpopulations [6,7].
Nitroaromatic antibiotics, such as metronidazole and nitrofurantonin, are widely used to treat anaerobic bacterial infections ( Figure 1). Based on this knowledge, researchers at the PathoGenesis Corporation developed a nitroimidazole pyran, PA-824 ( Figure 1), with specific anti-tuberculosis activity against both actively growing and hypoxic tuberculosis bacilli [8]. PA-824 is structurally related to metronidazole and was selected for its specificity in terms of antitubercular activity, favorable metabolic stability, low toxicity profile, and lack of cross-resistance with other TB drugs. PA-824 has advanced into Phase II combination clinical testing by the Global Alliance for TB Drug Development [8,9,10,11]. Researchers at Otsuka Pharmaceutical Company subsequently developed the nitroimidazole oxazole OPC-67683 ( Figure 1), which is now being progressed to Phase III clinical trials against tuberculosis [12,13,14]. These studies highlight the potential for the successful development of antitubercular nitroaromatics.
In addition to the nitroimidazoles, other nitroaromatics belonging to nitrothiazole and nitrofuran classes have notable antitubercular activity [8,12]. This includes our recently reported series of tetracyclic nitrofuranyl isoxazolines, which had improved solubility and bioavailability but decreased potency and metabolic stability compared to precursor nitrofurans [15,16]. Key differences in these compounds arise from differences in their redox potential, which drive how these compounds are activated and the types of reactive products that are produced resulting in different modes of cell killing [17]. Therefore all three classes of nitroaromatics have the potential to be simultaneously developed as TB treatments, as shown in this study. Previously, incorporation of the outer ring elements of OPC-67683 into tetracyclic nitrofuranyl isoxazolines produced a novel pentacyclic hybrid nitrofuran with improved metabolic stability but limited oral bioavailability, Lee1106 (herein referred to as compound 9a) [18]. Herein, we report the synthesis and evaluation of 9a and a series of 9a-analogs with improved in vivo exposure profiles (area under the curve, AUC). Of the pentacyclic hybrid nitrofurans reported, 9a was the most metabolically stable and did not show any adverse effects in vivo. Therefore, 9a remained the most promising lead candidate and was subjected to a detailed antimicrobial characterization. 9a was highly active in two separate in vitro models of nonreplicating tuberculosis, was active in a murine model of acute tuberculosis infection, and was activated by a mechanism distinct from that required for PA-824 and OPC-67683.

Reagents and Instrumentation
All anhydrous solvents and starting materials were purchased from Aldrich Chemical Co. (Milwaukee, WI). All reagent grade solvents used for chromatography were purchased from Fisher Scientific (Suwanee, GA) and flash column chromatography silica cartridges were obtained from Biotage Inc. (Lake Forest, VA). The reactions were monitored by thin-layer chromatography (TLC) on pre-coated Merch 60 F 254 silica gel plates and visualized using UV light (254 nm). A Biotage FLASH column chromatography system was used to purify mixtures. All NMR spectra were recorded on a Bruker-400 spectrometer. Chemical shifts (d) are reported in parts per million relative to the residual solvent peak or internal standard (tetramethylsilane), and coupling constants (J) are reported in hertz (Hz). High resolution mass spectra were recorded on a Waters Xevo G2 QTOF LCMS using ESI. Purity of the products was confirmed before testing by analytical RP-HPLC on a Shimadzu HPLC system, and all final compounds had a purity of 95% or greater as determined by RP-HPLC. Gradient Conditions: solvent A (0.1% formic acid in water) and solvent B (0.1% formic acid in acetonitrile): 0-1.00 min 95% A, 1.00-6.00 min 0-95% B (linear gradient), 6.00-9.50 min 100% B, 9.50-9.75 min 0-95% A, 9.75-10.0 min 95% A, detection by UV at 254 nm and by ELSD. The reference compound PA-824 was synthesized in six step sequence following the procedure reported by Baker et al., starting from 2,4-dinitro, 1H-imidazole and Sglycidol [19]. OPC-67683 was synthesized from 2-chloro-4-nitro-1H-imidazole as described by Tsubouchi et al. [20].

General Procedure II, for Preparation of Compounds 6a-b
Compounds 5a-b were dissolved in 1:1 CH 2 Cl 2 and trifluoroacetic acid (TFA) and stirred at room temperature for 1 h and concentrated under reduced pressure, washed with water, 10% NaOH and water, dried over anhydrous Na 2 SO 4 and concentrated to afford amines 6a-b in 92-95% yields.

General Procedure IV, for Preparation of 9a-f
At room temperature, with vigorous stirring, a solution of Et 3 N (1.2 mmol) in anhydrous CHCl 3 was slowly added to a solution of olefin (1.0 mmol) and N-hydroxy-5-nitrofuran-2-carbimidoyl chloride 8 (1.2 mmol) in anhydrous CHCl 3 . The reaction mixture was stirred at room temperature for 2 h then diluted with excess CHCl 3 , washed with water, dried over anhydrous Na 2 SO 4 , concentrated under reduced pressure and purified by flash chromatography to afford compounds 9a-i in 50-68% yields.

MIC Determination
Minimum inhibitory concentrations (MICs) were determined according to Clinical Laboratory Standards Institute (CLSI), using microbroth dilution method in 96-well plates and read by visual inspection [21]. The lowest concentration of drug that prevented visible growth was recorded as the MIC. Liquid MICs for indicated mycobacterium were determined in Middlebrook 7H9 broth supplemented with 10% albumin-dextrose complex and

Minimum Bactericidal Concentration Determination
Liquid MICs were performed as described above with the additional step of determining the number of bacteria added to each well. Serial dilutions of bacteria from the MIC plates were prepared and spread on Middlebrook 7H11 agar supplemented with 10% oleic acid-albumin-dextrose complex. After 4 weeks of incubation at 37uC, colony forming units (CFUs) were counted and the input calculated. Following 7 days of incubation, the contents of individual MIC plate wells were resuspended and spread on drug-free 7H11 agar plates. Plates were incubated at 37uC for 4 weeks prior to enumeration of viable colonies. The MBC was recorded as the lowest treatment concentration that killed .99.9% of input bacteria.

Cytotoxicity
In vitro cytotoxicity was assessed using Vero cells (kidney epithelial cells; ATCC CCL-81) essentially as described previously, except that detection was performed using CellTiter-GloH Luminescent Cell Viability (Promega) [22]. The concentration of treatments that produced a 50% decrease in viability (IC 50 values) were computed using nonlinear regression assisted equation of log [inhibitor] vs. Response-variable slope (four parameters) symmetrical, in GraphPad prism software.

Post Antibiotic Effect (PAE)
PAE studies were performed essentially as described previously [21]. To facilitate sample handling, M. bovis strain BCG Pasteur was used as a surrogate for M. tuberculosis [21]. Briefly, a mid-log phase culture recovered from frozen stock was diluted to an OD 600 of ,0.001 and allowed to grow to early log-phase (OD 600 0.2). For each experimental condition, 50 mL of cells were exposed to 9a (at 0.1, 1.0, or 10.0 mg/mL) or isoniazid (at 0.05, 0.5, or 5.0 mg/ mL) for 2 hours. Drugs were subsequently removed by washing twice in pre-warmed Middlebrook 7H9 medium and treated cells resuspended in pre-warmed medium. OD 600 was determined for each culture immediately following resuspension and every 24 hours thereafter until reaching growth saturation. The difference between the time required for treated cells to reach 50% of the maximum OD 600 of the untreated culture and for the untreated culture to reach this density was calculated as the PAE. Average and standard deviation were calculated based on results of two biologically independent experiments.   (1)

In vitro Activity against NRP Bacteria
To evaluate the activity of compounds against dormant TB bacilli, two in vitro models were used, mimicking oxygen and nutrient starvation. These experiments were performed using M. tuberculosis strain H37Rv as described previously, with no modifications [23,24]. Briefly, hypoxic cultures were exposed to 10 mg/mL of indicated treatments for 4 days prior to cfu enumeration. For nutrient starvation experiments, aliquots of six-week starved cultures (in PBS) were exposed to indicated treatments for one week prior to cfu enumeration.

Checkerboard Synergy Assays
Whole cell in vitro synergy assays were performed using M. tuberculosis strain H37Rv. In a 96-well assay plate, two fold serial dilutions of Drug A (ethambutol, isoniazid, linezolid, PA-824, rifampin, or streptomycin) were prepared in 100 ml of Middlebrook 7H9 media (highest and lowest concentrations in rows A and G, respectively, with no drug in row H). Using a single dip with a 200ss pintool, 0.2 ml of drug B (9a) was transferred to the assay plate columns 1 to 11 of the assay plate, with drug-free DMSO transferred to column 12. To each well of the assay plate 100 ml of mid-log phase bacteria (diluted to OD 600 of 0.01) was added, and plates incubated for 7 days prior to reading MICs by visual inspection.

In vitro Drug Resistance Frequency
Resistance frequency determination and spontaneous mutants selection was performed by plating M. tuberculosis H37Rv in the presence of increasing super-MIC concentrations of compound as described previously [21]. Where indicated, liquid cultures were exposed to a sub-inhibitory concentration of 9a prior to selection on agar.

Pharmacokinetic Analyses
Maximum solubility, metabolic stability, and plasma protein binding assays were performed and analyzed as described previously [22]. For in vivo pharmacokinetic studies, catheterized male Sprague-Dawley rats were obtained from Harlan Bioscience (Indianapolis, IN) and maintained as previously described [22]. Test compounds were dissolved in 30% DMSO, 30% propylene glycol, 20% PEG 3000 and 20% saline and administered intravenously (IV) at 10 mg/kg or were dissolved in PEG-3000:Water (60:40) and dosed by oral gavage at 100 mg/kg. Blood sample collection, processing, and analysis were performed exactly as described previously [22]. The pharmacokinetic parameters maximum plasma concentration (C max ), minimum plasma concentration within 24 hours after dosing (C min,24 h ), systematic exposure (AUC 0-' ), half-life (t 1/2 ), clearance (CL), volume of distribution (Vd), fraction excreted unchanged in urine (f e ), and oral bioavailability (F) were determined by standard noncompartmental analysis using the software package Phoenix WinNonlin 6.2 (Pharsight, Mountain View, CA).

In vivo Efficacy
Efficacy of 9a in an acute tuberculosis infection mouse model was performed as described previously [26,27,28]. Briefly, 8 week  ). Groups of 5 mice per group received 9a (300 mg/kg), isoniazid (25 mg/kg), or no treatment (untreated control group) for 9 consecutive days by oral gavage prior to sacrifice. Lungs and spleens from drug treated and control groups were homogenized and bacterial loads determined by CFU enumeration to determine bacterial loads before and after the 9 day treatment period, as described previously [27]. Reduction in bacterial loads of treatment groups compared to the untreated control group was recorded as (Log 10 ) reduction. For statistical analysis the CFU were converted to logarithms, which were then evaluated by a one-way analysis of variance and two-way analysis of variance, followed by a multiple comparison analysis of variance by a one-way Tukey test and Dunnett test (SAS Software program, Research Triangle Park, NC). Differences were considered significant at the 95% level of confidence.

Ethics Statement
The pharmacokinetic study protocol was approved by Institutional Animal Care and Use Committee of the University of Tennessee Health Science Center (Protocol number 1463). The in vivo efficacy study protocol was approved by Colorado State University Institutional Animal Care and Use Committee (Protocol number 13-4263A).

Synthesis of an Expanded Set of Pentacyclic Nitrofurans
The design of the target series of new pentacyclic nitrofurans is shown in Figure 1. The nitrofuranyl and isoxazoline rings of our previous lead compounds 1, 2 and 9a were kept intact to maintain antitubercular potency, whereas the outer side chain (C-E rings) of the pentacyclic system were modified [18,22]. Within this general scaffold, modifications were targeted to the C and E rings with substitutions aimed at blocking sites of potential metabolism and increasing solubility. The target compounds 9a-f were synthesized in a five step sequence starting from 1-Boc-4-hydroxypiperidine 3 as shown in Figure 2. Accordingly, 3 was mesylated by treating with methanesulfonyl chloride and triethylamine in dichloromethane at 0uC for 1 h to yield 4 in 93% yield which upon treatment with appropriate substituted E-ring phenols in presence of potassium carbonate and tetra n-butyl ammonium chloride in water at reflux afforded ethers 5a,b in 86-90% yields. Boc deprotection of 5a,b with trifluoroacetic acid in dichloromethane at room temperature produced the key intermediate amines 6a,b. The aryl amination reactions were carried out under Buchwald conditions by using appropriate C ring aryl bromides and secondary amines 6a,b in presence of palladium catalyst and sodium tert-butoxide in toluene at 100uC to afford 7a-f in 50-70% yields [22]. Finally, the isoxazoline ring was constructed by treating the olefins with nitrofuranyl chloroxime 8 [29] in the presence of triethylamine in chloroform at room temperature to give compounds 9a-f in 50-68% yields.

Antituberculosis Activity and in vitro Metabolic Profile
All the compounds in the series maintained excellent potency against M. tuberculosis (MIC 0.001-0.02 mg/mL, corresponding to 1.9-46 nM), as shown in Table 1, which was well within their C max range. Mammalian cytotoxicity assays and calculation of subsequent therapeutic indices revealed very favorable selectivity, which ranged from 150-2500. Solubility of the series was low, however, with modifications yielding a ,10 fold drop in solubility compared to our earlier series (2). Plasma protein binding remained very high (.96%) for all analogs. Compounds with a 4-fluorophenyl E ring (9d-f) showed a slight improvement in microsomal stability (t K of 1.4-1.9 h. compared to 1.1 h for 2). The 4-trifluoromethoxyphenyl E ring series (9a-c), however, demonstrated much improved metabolic stability (t K of 5-7 hours). Modification to the C ring (pyridyl in 9b and 9e, fluorophenyl in 9c and 9f) had negligible effects on overall properties irrespective of E ring substitution.

Spectrum of Antimicrobial Activity
Compounds 9a-f were weakly active or inactive against a panel of gram-positive and gram-negative pathogens (Table S1) except for 9f, which had moderate activity against most gram-positive pathogens and an efflux pump deficient strain of E. coli. This activity is consistent to that observed for other similar nitrofurans we generated in prior series [21,22]. Compound 9a was further tested against non-tuberculosis disease causing mycobacteria   [30,31]. The activity for 9a against M. abscessus is notable since this difficult to treat mycobacterium is intrinsically resistant to most anti-mycobacterial agents [32]. These differences in the spectrum of activity between the nitrofurans and the nitroimidazole PA-824 is consistent with differences in the redox potential for these two compound classes, with the lower redox potential of nitroimidazoles requiring specific enzymes for activation, whilst nitrofurans have a higher redox potential and may be more easily activated by a wider spectrum of enzymes [17,33].

In vivo Pharmacokinetic Profiling
The pharmacokinetic profiles of 9b-c were evaluated in rats and compared to the published data for 9a using similar formulations (Table 3). After intravenous administration of 10 mg/kg 9b-c, compounds showed biexponential concentration-time profiles similar to 9a, but elimination half-lives (tK 2.5-5.6 h) significantly lower than 9a (tK 10.3 h). The shorter halflives are a consequence of a reduced volume of distribution for compounds 9b-c (Vd 1.53-1.69 L/kg) compared to 9a (Vd 6.72 L/kg), while clearance (CL) remained similar for all three compounds. The fraction of dose excreted unchanged by the kidneys (fe) was negligible for all compounds. Their relatively large volumes of distribution and non-renal elimination are probably a reflection of their high lipophilicity (clogP 4.6-5.8). The oral bioavailability of compounds 9b and 9c was poor (F 8.0%, 0.83%), possibly a reflection of poor solubility as had already been noted for 9a which also has limited oral bioavailability (F 4.6%).Compounds 9a and 9b were non-toxic by intravenous and oral dosing in rats. Central nervous system toxicity (seizures) was observed for compound 9c in both administration modes. Compound 9a was therefore advanced for in vivo efficacy testing due to its longer in vivo half-life and higher volume of distribution as compared to 9b.
In vivo Efficacy in a Mouse Model of Acute Tuberculosis Infection 9a was next evaluated for in vivo efficacy using a mouse model of acute tuberculosis infection [26]. Given the poor solubility of 9a, six different formulations were employed in the acute GKO mouse infection model. In each formulation, 9a was well tolerated with daily oral administration at 300 mg/kg and 9 days of treatment provided a statistically significant reduction (P,0.001 by Tukey Test) in bacillary loads in both the lungs and spleen (Figure 3, Table S2). The differences in efficacy between different 9a formulation groups were not statistically significant. A control drug isoniazid was included in the experiment (data not shown), which gave an expected a 3.5 and 4.6 log reduction of the bacterial load in lungs and spleen, respectively, versus the untreated controls. Although 9a did not provide superior protection compared to rapid acting control isoniazid, the reduction it produced in lungs was 91-97% in 9 days. This marks a substantial improvement over precursor tetracyclic nitrofuranyl isoxazolines, which were not active in vivo [15].

Post Antibiotic Effect and in vitro Synergy with Frontline Agents
Post antibiotic effect (PAE), defined as the duration of growth retardation following the removal of an antibiotic from the environment of the organism, is an important parameter for early consideration of potential dosing and was determined for compound 9a [34]. Pulse exposure to 9a yielded a doseresponsive PAE, with compound concentrations of 0.01, 1.0, and 10.0 mg/mL yielding PAE's of 21.7+/21.2 h, 105.5+/2 19.8 h, and 178.4+/28.8 h, respectively. The PAE produced by treatment with 9a at 1 mg/mL (105.5 h) was superior to that produced by exposure to 5 mg/mL isoniazid (71.8 h). The long PAE of compound 9a is highly favorable for the further development of this antitubercular compound series.
An important characteristic for novel tuberculosis therapeutics is compatibility with antitubercular agents in the clinic and late stage development, since combination therapy is applied to reduce the development of acquired drug-resistance. A preliminary investigation of 9a's potential for co-administration was, therefore, conducted using the checkerboard approach ( No synergy or antagonism was seen with frontline drugs ethambutol and isoniazid, nor with linezolid or streptomycin (FICI scores 0.5-4.0). However, 9a synergized with the frontline treatment rifampin and nitroimidazole PA-824 (FICI ,0.5). Interestingly, in the presence of 9a, MICs of PA-824 and rifampin both decreased ,10 fold. The favorable synergy predicted for 9a with rifampin, as well as the lack of unfavorable interactions predicted with other antitubercular drugs, suggests that 9a may complement current tuberculosis treatment regimes.

Minimum Bactericidal Concentration Determinations
The minimum bactericidal concentration (MBC) of 9a was determined (alongside controls); the results are shown in Table 5. The number of viable bacteria in individual wells of MIC plates was enumerated to determine the minimum bactericidal concentration (MBC), defined as the treatment concentration that killed .99.9% of the inoculum. Streptomycin was cidal (MBC/MIC #4), as anticipated, whereas all nitroaromatics compounds investigated were static (MBC/MIC .4). These observations are also consistent with nitrofurans being known to be static at low concentrations but cidal at higher concentrations.

Activity against Multiple Models of Tuberculosis
Several in vitro models have been proposed to mimic the physiological state of persistent tuberculosis infections in vivo. The most commonly used model of tuberculosis persistence is based upon the assumption that the host immune system sequesters bacilli into granulomas, where they persist at low oxygen levels. Activity of 9a was assessed in the Rapid Anaerobic Dormancy (RAD) model which uses rapid oxygen depletion to mimic the Non-Replicating-Persister 2 (NRP-2) conditions described by Voskuil. [24,36]. As anticipated, NRP-2 bacilli were sensitive to treatment with rifampin but resistant to isoniazid (Table 6). In this in vitro model of a dormant tuberculosis, 9a and PA-824 were both highly active, killing .99.9% of hypoxic bacilli at equivalent concentrations (10 mg/mL). In addition to oxygen deprivation, nutrient starvation is also proposed as an important environmental inducer of persistent tubercular populations. A simple model of starvation-induced persistence reported by Betts et al. [23]was adopted to further assess the anti-dormancy activity of nitrofuran 9a (Figure 4). Mid-log and nutrient-starved cultures were treated with test compound (9a or isoniazid at 1 mg/mL) or DMSO drug carrier for 7 days. As anticipated, isoniazid treatment killed .95% of actively growing M. tuberculosis culture but was 7 fold less active against nutrient starved bacteria. 9a produced an 80% reduction in viability of both actively growing and nutrient starved bacilli. The sustained activity of compound 9a in two separate models of biologically relevant, drug-refractory M. tuberculosis populations suggests it may serve as a valuable, treatment-shortening addition to the existing anti-tuberculosis regimen.

Mechanism of Action Studies
All nitroaromatic antibiotics are prodrugs for which bioreduction by the pathogen is required to generate the active antimicrobial metabolites. The bioreduction of PA-824 is well studied and is mediated by the deazaflavin-dependent nitroreductase (Ddn) in M. tuberculosis, with cellular bioreduction ultimately producing toxic nitric oxide [37,38]. The F 420dependent glucose 6-phosphate dehydrogenase (FGD1), which is required in the recycling of F 420 cofactor, is responsible for transferring electrons to Ddn (encoded by Rv3547). Thus mutations in genes encoding Ddn, FGD1 and F 420 biosynthesis confer resistance to PA-824 and cross resistance to OPC-67683, which is activated in a similar manner [39]. A collection of PA-824 resistant mutants each harboring a mutation in a single gene in the PA-824 activation pathway (FGD1, F 420 , or Ddn) were screened against compound 9a for cross resistance ( Table 7). All these mutants remained sensitive to 9a, suggesting that 9a is primarily activated by a different pathway and is unlikely to be affected by primary resistance mechanisms of bicyclic nitroimidazoles.
To assess M. tuberculosis's potential to become resistant to lead 9a, the frequency at which spontaneous resistance arises was determined by growing the susceptible M. tuberculosis strain H37Rv on agar containing 9a. Resistant clones arose at a frequency of 3610 27 in the presence of 2-46MIC of 9a. Repeated efforts to generate resistant clones at higher concentrations of 9a were unsuccessful, suggesting the potential for development of resistance is very low, possibly attributable to multiple mechanisms of activation or the induction of alternative modes of action at higher concentrations of compound. This appears consistent with 9a being bactericidal at higher concentrations. Pre-treatment of cultures with sub-inhibitory concentrations of 9a permitted selection of resistant colonies on agar plates containing 0.05-0.4 mg/mL of 9a. All 9a resistant clones assessed exhibited elevated 9a MIC (.3.13 mg/mL) but retained sensitivity to ethambutol, linezolid, streptomycin, rifampin, ofloxacin, and isoniazid (Table 8). All but a single 9a resistant clone remained sensitive to PA-824 and OPC-67683 (Table 8) further supporting the hypothesis that 9a is primarily activated by a unique biological pathway.

Conclusions
Given the success of nitroaromatic antibiotics in treating anaerobic bacterial infections and the current lack of suitable therapeutics to treat persistent tuberculosis infections, we report here the synthesis and antimicrobial characterization of pentacyclic nitrofuran analogs that incorporate the outer ring elements of OPC-67683. These pentacyclic nitrofurans displayed a favorable in vitro selectivity index and selective inhibitory activity against mycobacteria. The MIC of the series lead 9a (Lee1106) was 46 nM, comparable to that of OPC-67683 (19 nM), and is 20-fold more potent than PA-824 (.1000 nM). Compound 9a was active in in vitro assays against nonreplicating bacteria suggesting its potential for treatment of chronic tuberculosis infections. 9a exerted a lengthy post antibiotic effect, similar to isoniazid, and was synergistic with the frontline drug rifampin. Mechanism of actions studies showed that the three genes required for activation of PA-824 and OPC-67683 are dispensable for 9a antitubercular activity, indicating a low potential for cross resistance to other nitroimidazole antitubercular agents. This finding also shows that despite the sharing of common outer ring structural features between 9a and OPC-67683, the key difference in their nitroaromatic head groups leads to the use of different mechanisms of bioreductive activation. This likely arises from significant differences in the redox potential between nitrofurans and nitroimidazoles, with the latter compounds having a lower redox potential which means that more specific enzymes are required to transfer electrons to the nitro-group. Conversely, the high redox potential for nitrofurans makes these molecules easier to be reduced and different enzymes may be involved [17,40]. Compound 9a was well tolerated in a murine model of acute tuberculosis infection, and showed significant in vivo efficacy in both lungs and spleens with .90% reduction in bacillary load after 9 days of treatment. In summary, the studies presented here show significant promise for development of the nitrofuran series as antituberculous agents. Further studies to improve metabolic stability, solubility and bioavailability are warranted in order to enhance the in vivo efficacy of the series.

Supporting Information
Table S1 Spectrum of Antimicrobial Activity. (DOCX)