Intracellular Accumulation of Novel and Clinically Used TB Drugs Potentiates Intracellular Synergy

ABSTRACT The therapeutic repertoire for tuberculosis (TB) remains limited despite the existence of many TB drugs that are highly active in in vitro models and possess clinical utility. Underlying the lack of efficacy in vivo is the inability of TB drugs to penetrate microenvironments inhabited by the causative agent, Mycobacterium tuberculosis, including host alveolar macrophages. Here, we determined the ability of the phenoxazine PhX1 previously shown to be active against M. tuberculosis in vitro to differentially penetrate murine compartments, including plasma, epithelial lining fluid, and isolated epithelial lining fluid cells. We also investigated the extent of permeation into uninfected and M. tuberculosis-infected human macrophage-like Tamm-Horsfall protein 1 (THP-1) cells directly and by comparing to results obtained in vitro in synergy assays. Our data indicate that PhX1 (4,750 ± 127.2 ng/ml) penetrates more effectively into THP-1 cells than do the clinically used anti-TB agents, rifampin (3,050 ± 62.9 ng/ml), moxifloxacin (3,374 ± 48.7 ng/ml), bedaquiline (4,410 ± 190.9 ng/ml), and linezolid (770 ± 14.1 ng/ml). Compound efficacy in infected cells correlated with intracellular accumulation, reinforcing the perceived importance of intracellular penetration as a key drug property. Moreover, we detected synergies deriving from redox-stimulatory combinations of PhX1 or clofazimine with the novel prenylated amino-artemisinin WHN296. Finally, we used compound synergies to elucidate the relationship between compound intracellular accumulation and efficacy, with PhX1/WHN296 synergy levels shown to predict drug efficacy. Collectively, our data support the utility of the applied assays in identifying in vitro active compounds with the potential for clinical development. IMPORTANCE This study addresses the development of novel therapeutic compounds for the eventual treatment of drug-resistant tuberculosis. Tuberculosis continues to progress, with cases of Mycobacterium tuberculosis (M. tuberculosis) resistance to first-line medications increasing. We assess new combinations of drugs with both oxidant and redox properties coupled with a third partner drug, with the focus here being on the potentiation of M. tuberculosis-active combinations of compounds in the intracellular macrophage environment. Thus, we determined the ability of the phenoxazine PhX1, previously shown to be active against M. tuberculosis in vitro, to differentially penetrate murine compartments, including plasma, epithelial lining fluid, and isolated epithelial lining fluid cells. In addition, the extent of permeation into human macrophage-like THP-1 cells and H37Rv-infected THP-1 cells was measured via mass spectrometry and compared to in vitro two-dimensional synergy and subsequent intracellular efficacy. Collectively, our data indicate that development of new drugs will be facilitated using the methods described herein.

ABSTRACT The therapeutic repertoire for tuberculosis (TB) remains limited despite the existence of many TB drugs that are highly active in in vitro models and possess clinical utility. Underlying the lack of efficacy in vivo is the inability of TB drugs to penetrate microenvironments inhabited by the causative agent, Mycobacterium tuberculosis, including host alveolar macrophages. Here, we determined the ability of the phenoxazine PhX1 previously shown to be active against M. tuberculosis in vitro to differentially penetrate murine compartments, including plasma, epithelial lining fluid, and isolated epithelial lining fluid cells. We also investigated the extent of permeation into uninfected and M. tuberculosis-infected human macrophage-like Tamm-Horsfall protein 1 (THP-1) cells directly and by comparing to results obtained in vitro in synergy assays. Our data indicate that PhX1 (4,750 6 127.2 ng/ml) penetrates more effectively into THP-1 cells than do the clinically used anti-TB agents, rifampin (3,050 6 62.9 ng/ml), moxifloxacin (3,374 6 48.7 ng/ml), bedaquiline (4,410 6 190.9 ng/ml), and linezolid (770 6 14.1 ng/ml). Compound efficacy in infected cells correlated with intracellular accumulation, reinforcing the perceived importance of intracellular penetration as a key drug property. Moreover, we detected synergies deriving from redox-stimulatory combinations of PhX1 or clofazimine with the novel prenylated amino-artemisinin WHN296. Finally, we used compound synergies to elucidate the relationship between compound intracellular accumulation and efficacy, with PhX1/WHN296 synergy levels shown to predict drug efficacy. Collectively, our data support the utility of the applied assays in identifying in vitro active compounds with the potential for clinical development. IMPORTANCE This study addresses the development of novel therapeutic compounds for the eventual treatment of drug-resistant tuberculosis. Tuberculosis continues to progress, with cases of Mycobacterium tuberculosis (M. tuberculosis) resistance to first-line medications increasing. We assess new combinations of drugs with both oxidant and redox properties coupled with a third partner drug, with the focus here being on the potentiation of M. tuberculosis-active combinations of compounds in the intracellular macrophage environment. Thus, we determined the ability of the phenoxazine PhX1, previously shown to be active against M. tuberculosis in vitro, to differentially penetrate murine compartments, including plasma, epithelial lining fluid, and isolated epithelial lining fluid cells. In addition, the extent of permeation into human macrophage-like THP-1 cells and H37Rv-infected THP-1 cells was measured via mass spectrometry and compared to in vitro two-dimensional synergy and subsequent intracellular efficacy. Collectively, our data indicate that development of new drugs will be facilitated using the methods described herein.
T uberculosis (TB) remains a leading cause of death from an infectious disease, surpassing both malaria and human immunodeficiency virus/acquired-immunodeficiency syndrome. The World Health Organization (WHO) estimates the total number of TB cases globally at 10 million, with 1.2 million annual deaths (1). The current rate of global TB decrease (1 to 2%) will not be sufficient to achieve the WHO's End TB goal of eliminating TB by 2035 (2). Moreover, the emergence of multidrug-resistant (MDR) and extensively drug-resistant (XDR) TB has exacerbated the scale of this challenge (1). Consequently, there is an urgent need for new anti-TB drugs and combination regimens to reduce treatment duration and combat drug-resistant disease.
Pulmonary alveolar macrophages and the epithelial lining fluid (ELF) have been identified as important sites of infection for numerous diseases, including TB, which is caused by Mycobacterium tuberculosis (M. tuberculosis). Following cellular uptake, M. tuberculosis can be sequestered into subcellular compartments, including phagolysosomes, lysosome, and cytosol (3,4), which further alter M. tuberculosis replication rates and susceptibility to antibiotic treatment (5,6). The sterilization of the intracellular environment should therefore be the goal of drug discovery programs (7)(8)(9)(10), particularly given the potential that novel drugs targeting specific intracellular bacilli may hold for shortening treatment (11)(12)(13)(14).
Tamm-Horsfall protein 1 (THP-1) is a spontaneously immortalized human monocyte-like cell line (15) which has been widely applied in investigations of intracellular M. tuberculosis infection (16,17). An advantage of this model is that macrophage-like cell lines display representative features of in vitro-differentiated monocyte-derived macrophages (18) while obviating the potential confounder of donor variability in macrophage function. In addition, THP-1 cells can be grown reproducibly, can be studied at different stages (e.g., resting versus activated), are easily infected, and closely model alveolar macrophages for M. tuberculosis-induced apoptosis (19).
The epithelial lining fluid has recently been identified as an important site to measure drug concentrations owing to its role as potential bacterial reservoir (20,21). Concentrations of antibiotic drugs in lung fluids, including sputum, respiratory tract secretions, bronchial mucosa, epithelial lining fluid, and bronchoalveolar lavage (BAL) fluid, have been measured using a number of different methods (22,23). Tissue samples are often homogenized, and while they provide a good average concentration of overall drug penetration into organs, information on specific compartmental penetration is lost owing to the processing of organ samples, sometimes leading to poor prediction of target-site drug concentration (24). A significant impediment to TB treatment is caused by the complexity of TB pathology, with the disease progressing from the crucial incipient macrophage infection through to lesion development (25). Drug concentrations in epithelial lining fluid and macrophages provide the best predictive values of target-site concentrations because of their proximity to the initial target site of mycobacterial infection. However, obtaining these samples is often difficult and uncomfortable for patients, involving the insertion of a bronchoscope and washing of the lung environment (26). Nevertheless, the BAL procedure allows for sampling of the epithelial lining fluid and the associated cellular fraction within this fluid, consisting predominantly of monocytic/macrophage cells (27) and providing an ideal in vivo comparator for our in vitro intracellular model.
In this study, we evaluated the intracellular accumulations and efficacies as well as potential synergizing activities of the phenoxazine derivative PhX1 (M. tuberculosis MIC 90 0.19 mM [28]), the prenylated artemisinin derivative WHN296 (MIC 90 18 mM [29]), the isoniazid-based semicarbazone DPINH (MIC 90 0.36 mM [30]), and the decoquinate derivative RMB041 (MIC 90 1.61 mM [31]), as presented in Figure 1. Clofazimine, which structurally resembles the phenoxazine PhX1, was included as a comparator drug. Both compounds are lipophilic and weakly basic. As such, these drugs, with their lysosomal trapping capacity, tend to accumulate in acidic organelle compartments via pH-dependent ion trapping mechanisms (32). Importantly, clofazimine is equipotent in vitro against both MDR and drug-sensitive M. tuberculosis, due to its activity on the M. tuberculosis type II NADH:quinone oxidoreductase (M. tuberculosis NDH2), where it induces the formation of reactive oxygen species via redox cycling that is associated with the increased consumption of NADH (33,34). Clofazimine-like compounds are ideal for inclusion in this regimen, as both MDR and drugsensitive TB patients could be treated using similar treatment regimens. In the context of the amino-artemisinin derivative WHN296, it is noted that an artemisinin derivative specially tailored to be taken up via an active transporter into M. tuberculosis (35) and a lipophilic steroidal tetroxane analogue (36) have previously been shown to be active against M. tuberculosis in vitro. More recently, amino-artemisinins and tetraoxanes bearing diamine linkers attached to a pyrimidine showed substantial growth suppression of M. tuberculosis cultures in vitro (37).
The examination of the anti-M. tuberculosis-active phenoxazines, semicarbazones, decoquinate derivatives, and amino-artemisinin derivatives in comparison with known TB drugs is therefore warranted in the special context of intracellular accumulation, as is now described.

RESULTS
PhX1 accumulation in murine bronchoalveolar lavage fluid. LC-MS/MS assay performance. A liquid chromatography tandem mass spectrometry (LC-MS/MS) assay was used for the analysis of PhX1 in infected THP-1 samples with calibration standard and quality control (QC) accuracy (%Nom) between 89.5 6 7.4 and 110.4 6 2.2%. This indicated that the calibration curve values for the murine BAL fluid, plasma, and alveolar macrophage samples were well within the acceptable 20% deviation used in this study for both calibration curve and QC values and that the LC-MS/MS assay performed well in the analysis of these murine samples.
PhX1 concentration in lung-associated fluids. To investigate whether PhX1 accumulated within different lung environments, a BAL procedure was conducted on mice treated with PhX1, with parallel plasma sampling conducted at corresponding time points, followed by LC-MS/MS analysis. This allowed for the determination of compound concentrations in the epithelial lining fluid, alveolar macrophages, and plasma (Fig. 2).
Samples were collected over a 24-h period, and the analysis indicated that PhX1 accumulated to a significantly higher concentration (Fig. 2) in the alveolar macrophages than in the plasma and epithelial lining fluid. PhX1 plasma concentrations in this study were greater than the M. tuberculosis MIC 90 value for this compound for approximately 6 h, achieving a maximum concentration (C max ) of 200 6 75 ng/ml. Drug exposure in plasma was calculated using the area under the concentration-time curve (AUC), providing a value of 1,172 ng/ml/h. The concentration of PhX1 in the epithelial lining fluid was significantly higher than that in the plasma, with a time above This disparity between the epithelial lining fluid and plasma is further highlighted by the lung fluid cell concentrations. The predominant cell type within the epithelial lining fluid is pulmonary alveolar macrophages (83%) (23), in which PhX1 accumulates to significant levels ( Fig. S1). The concentrations of compound measured in the epithelial lining fluid cells were significantly higher than those recorded in plasma and epithelial lining fluid, achieving a C max of 895 6 23 ng/ml with a calculated AUC of 5,018 ng/ml/h, approximately 4-fold higher than that in plasma samples and 2-fold higher than that in the epithelial lining fluid.
Distinctive accumulation ratios reported in uninfected THP-1 cells over a 96-h period. LC-MS/MS assay performance. A quadratic regression equation, plotting peak area ratio against concentration, was fitted to the calibration curves. The curves were weighted by 1/concentration (1/x). The accuracy (%Nom) for all calibration standards and QC samples was between 91.2 6 2.5% and 109.1 6 4.9% in this study. This indicated that calibration curve and QC values were well within the acceptable 20% deviation used in this study for the THP-1 LC-MS/MS cell samples.
Further investigation into the macrophage-accumulation of PhX1 was conducted using an in vitro THP-1 cell model and treated with PhX1 and other experimental and approved TB drugs over a period of 48 h (Fig. 3A). Cell number and viability were unaffected by the addition of drug for the period of 48 h, with cell numbers maintained at values greater than 4 Â 10 5 cells/well for all samples. PhX1 displayed the greatest level of intracellular accumulation (4,750 6 127.2 ng/ml), with bedaquiline (4,410 6 190.9 ng/ml), moxifloxacin (3,374 6 48.7 ng/ml), and rifampin (3,050 6 62.9 ng/ml) displaying relatively high intracellular compound levels (Fig. 3B). Lower intracellular concentrations were noted for WHN296 and RMB041 and for the known drugs linezolid and levofloxacin (Table 1). Cell viability remained consistent over the course of the experiment, with viability greater than 85% for all samples (Table 1). Statistical analysis revealed that PhX1 accumulated to significantly higher levels than the experimental compounds and several of the clinical drugs included in this study.
Phenoxazine accumulation in Mycobacterium smegmatis-infected THP-1 cells. To investigate whether intracellular accumulation of PhX1 was related to compound structure, several phenoxazine derivatives were assessed for their ability to accumulate in M. smegmatis-infected THP-1 cells. M. smegmatis was used as a bench-safe alternative to M. tuberculosis H37Rv, allowing the development of an assay measuring intracellular accumulation and other readouts in an infection-based system. Multiple samples were taken to assess the time-dependent nature of this accumulation, but no change was seen following 1 h of compound incubation.
PhX1 accumulated to levels significantly higher than those of the majority of the other phenoxazines ( Fig. 4A and B), with PhX6 being the only compound accumulating to higher levels within this model. To visualize compound accumulation, we conducted fluorescence imaging on M. smegmatis-infected THP1 cells treated with PhX1 or phosphate-buffered saline (PBS)/ dimethyl sulfoxide (DMSO) (untreated) as shown in Figure 4C and D. Treatment with PhX1 resulted in intense red fluorescence, with comparatively less fluorescence seen in the untreated sample, indicative of accumulation within these cells ( Fig. 4C and D). However, no significant change in M. smegmatis-associated green fluorescence could be seen following PhX1 treatment, possibly indicating a lack of efficacy against M. smegmatis as reported for other structurally related compounds (38). This accumulation was confirmed using flow cytometry, with the uninfected and infected THP-1 cells accumulating PhX1 as indicated by the population movement across the allophycocyanin (APC) channel (Fig. 4E).
M. tuberculosis H37Rv-infected macrophage model. The approach used in this study focused primarily on a standardized 20Â MIC compound dose comparison to incorporate the assay into our screening cascade. This translated into concentrations similar to those achieved in the plasma and epithelial lining fluid for a number of the compounds.
Intracellular drug accumulation in M. tuberculosis-infected THP-1 cells drives efficacy. An intracellular accumulation ratio, defined as the intracellular compound concentration divided by the extracellular concentration, was used in this study to allow comparison between compounds (Fig. 5B). This required the measurement of both intracellular and extracellular concentrations of drug as shown in a number of publications (39)(40)(41)(42)(43)(44)(45)(46)(47). RMB041 showed the lowest levels of intracellular accumulation at 0.25, PhX1 displayed accumulation levels of around 2.5 to 3 (similar to rifampin), and moxifloxacin and bedaquiline showed the greatest accumulation in the infected THP-1 cells (5.4 and 5.5, respectively). The known drugs rifampin (41,43), moxifloxacin (42,43,48), and bedaquiline (49,50) have been shown to achieve similar intracellular accumulation ratios in other studies, validating these results. Additional efficacy measurements were obtained by plating for survival of M. tuberculosis H37Rv (Fig. 5C and D). The untreated control displayed an increased bacterial count of 1.6 log 10 CFU/ml compared to the starting bacterial count. In order of increasing efficacy, RMB041 displayed a decreased bacterial count of 0.1 log 10 CFU/ml, moxifloxacin (20.5 log 10 CFU/ml), PhX1 (20.8 log 10 CFU/ml), bedaquiline (21.5 log 10 CFU/ml), and rifampin (21.5 log 10 CFU/ml). Synergistic interaction in the M. tuberculosis-infected THP-1 model. A standard two-dimensional (2D) checkerboard/synergy assay was performed to investigate potential synergism between the compounds ( Fig. 6A and C). Drug interactions were assessed by checkerboard dilution in a 96-well format with slight modification of the method described by Omollo et al. (51). PhX1 was serially diluted in RPMI medium along the x axis (columns 3 to 11), with synergistic concentrations ranging from 25 to 0.19 mM, while WHN296 was serially diluted across the y axis (rows B to H), with synergistic concentrations ranging from 100 to 0.15 mM.
The PhX1/WHN296 combination displayed a fractional inhibitory concentration index (FICI) of 0.21 (Fig. 6A), which is defined as synergistic. The minimum values required to produce synergy were also noted for each compound, including PhX1 (153.1 ng/ml) and WHN296 (1,527.2 ng/ml). The results compared well with those obtained from the clofazimine/WHN296 combination, which also produced an FICI of 0.32 (Fig. 6C).
It was necessary to confirm if synergism between PhX1 and WHN296 was conserved in M. tuberculosis-infected THP-1 cells. To compare these results to the single-dose efficacy and combinations obtained using the individual drugs, each compound was dosed at 20Â MIC 90 ( Fig. 6B and D). The intracellular accumulation of the combination of PhX1 and WHN296 in the M. tuberculosis-infected THP-1 cell model was evaluated using quantitative LC-MS/MS at time points of 24, 48, 72, and 96 h (n = 3 for each time point). In this experiment, intracellular drug concentrations are reported in ng/ml to allow comparisons to be made with the minimum compound levels required to achieve synergy. For the PhX1/WHN296 combination, both compounds reached their respective intracellular synergy levels after approximately 48 h and maintained these levels until the 96-h time point.
Next, it was important to establish whether both drugs in the PhX1/WHN296 combination reaching these intracellular 2D levels resulted in a corresponding increase in

DISCUSSION
The success of M. tuberculosis as an infectious agent derives from its ability to persist within the human host (52). One of the most distinct features of M. tuberculosis is its ability to infect the macrophage, recently summarized by VanderVen et al. as the "minimal unit of infection" (53). This description encompasses both the innate immune response toward M. tuberculosis and the pathogenicity of M. tuberculosis. In addition, Dartois et al. eloquently outlined the need to understand how individual compounds distribute into the different host cell environments (8). It is generally acknowledged that in order to predict how a drug will respond in the human body, the pharmacokinetic/pharmacodynamic relationship in various tissues and cells must be understood (10,54,55). Given the propensity of M. tuberculosis to occupy different microenvironments of the body, it is essential that efforts are focused on developing drugs that are active in all environments.
This was addressed by collecting BAL fluid and plasma from uninfected C57BL/6 mice dosed with compounds over 24 h. Within the context of a therapeutic window, the importance of determining target-site concentrations becomes clearer (55). The evaluation of plasma concentrations in this study would enable us to establish if concentrations are maintained above the MIC for the compounds over a maximum of 7 h. However, for concentrations in the epithelial lining fluid cells, the time during which compound concentrations are above the MIC 90 would be closer to 15 h.
The ratio between epithelial lining fluid and plasma concentrations is a commonly used predictor of penetration into the target site and is usually presented as a log ratio, with a higher log ratio indicative of higher concentrations in epithelial lining fluid compared to those in the plasma (56,57). This log ratio for PhX1 (0.65) was higher than that for ciprofloxacin (58-60) at a ratio of 20.2 and similar to isoniazid (61) at a ratio of 1, linezolid (62)(63)(64) at 0.7, and rifampin (65,66) at 0.7, while ethionamide (67) at 2.8, moxifloxacin (68) at 1.5, and pyrazinamide (69) at 3.2 showed significantly higher ratios, indicative of their superior intrapulmonary penetration. However, recent literature points to several flaws associated with determining an epithelial lining fluid/plasma concentration using single-point values such as the C max (57,70,71). The systemic hysteresis associated with both epithelial lining fluid and plasma compound concentrations requires a more stable pharmacokinetic (PK) parameter to be determined, such as an AUC measurement. In this study, the ELF AUC / plasma AUC for PhX1 provided a value of 1.7, which still indicated moderate to high penetration and subsequently high level of epithelial lining fluid exposure compared to those of the clinically used compounds.
Given the cellular accumulation of PhX1 in vivo, further investigations were conducted using an in vitro THP-1 cell model and compared to intracellular accumulation of known drugs and new compounds, with PhX1 accumulating to greater levels than the majority of other compounds. To determine whether this was related to a structurally distinct feature of PhX1, several phenoxazine derivatives were tested in an M. smegmatis-infected THP-1 cell model. PhX1 displayed higher intracellular accumulation than the majority of the other phenoxazines. Given the 2-fold increase in PhX1 efficacy against in vitro M. tuberculosis over PhX6, it was decided that PhX1 would be followed in subsequent experiments. PhX1 LC-MS/MS results were further complemented using flow cytometry, which demonstrated the clear accumulation of PhX1 over time in the APC channel. This supported the accumulation seen in the fluorescence microscopy experiments using untreated and PhX1-treated M. smegmatis-infected cells. Drug accumulation over time was recorded using flow cytometry and LC-MS/MS analyses, with a steady state being achieved for both after 2 h.
Evaluation of patterns of drug accumulation within immune cells involved in granuloma formation, including macrophages, indicates that compound accumulation is not straightforward (8,25,72). Four distinct patterns of drug penetration have been identified, including (i) rapid and homogeneous distribution with the absence of accumulation appearing over time (isoniazid/linezolid), which may explain the predominant killing of extracellular bacteria by isoniazid, (ii) rapid and heterogeneous distribution with a high degree of accumulation in the cellular rim rather than the caseum (fluoroquinolones), (iii) slower distribution with gradual accumulation of drug over time (rifampin), and (iv) rapid distribution with significant accumulation in the cellular layers and poor diffusion into the caseum (clofazimine and bedaquiline). Rapid homogeneous distribution followed by a lack of accumulation is associated with linezolid, with an equilibrium level being reached within 15 min. It also has been shown elsewhere that linezolid displays intracellular concentrations lower than those of the majority of other compounds (73,74). The fluoroquinolones accumulate within a variety of cells, including THP-1 cells (75)(76)(77). Here, the uptake of the fluoroquinolones moxifloxacin and levofloxacin conformed to the rapid and heterogenous pattern of accumulation, with moxifloxacin accumulating to a slightly higher degree than levofloxacin. Others have compared the degree of accumulation of several fluoroquinolones at steady state (78), which was reached at approximately the same time as in our study (30 min). The frontline anti-TB drug rifampin accumulated exceedingly slowly but reached significant levels in the cellular regions of the granuloma. Rifampin attained high concentrations within THP-1 cells. Bedaquiline has been shown to attain high levels of accumulation within the cellular rim of granulomas and in THP-1 cells (79). Here, it displayed the same degree of accumulation, with equilibrium levels greater than 4 mg/ml being reached within 30 min. The compounds PhX1, DPINH, RMB041, and WHN296 (Fig. 1) displayed various degrees of accumulation. PhX1 showed relatively high levels of accumulation, comparable with the structurally similar clofazimine, which accumulates to high levels within various models (8, 72, 80). The accumulation levels measured for PhX1 were also similar to those of bedaquiline.
Comparatively lower accumulation was observed for DPINH. As noted above, this could be because of the potential of the compound to undergo metabolism to isoniazid, a compound with relatively poor cellular accumulation properties (7,49). RMB041 and WHN296 accumulated to relatively lower levels within uninfected THP-1 cells, achieving levels similar to those of linezolid and levofloxacin. These differential levels of accumulation were encouraging and allowed us to move forward with the implementation of further perturbations to the model.
M. tuberculosis is a facultative intracellular organism that is able to infect and grow in the macrophage environment (53,81,82). The infected THP-1 model is not unique and has been employed using M. smegmatis, Pseudomonas aeruginosa, and various other organisms, including M. tuberculosis (16,19,(83)(84)(85). However, very few studies have been conducted using M. tuberculosis H37Rv as the infectious agent while examining compound accumulation and intracellular efficacy within the same experimental system. The infected macrophage assay was developed based on the uninfected assay used in this study and according to methods described in several publications (8,43,49,84,(86)(87)(88)(89)(90). The approach utilized a standardized 20Â MIC compound dose comparison to incorporate the assay into our screening cascade. These concentrations presented clinically relevant BAL fluid and plasma concentrations for the majority of the compounds. Rifampin achieves maximum concentrations of approximately 2 mg/ml in the epithelial lining fluid of patients, a dose similar to that used in this study (26,91). The concentration of moxifloxacin at 5 mg/ml in plasma at the 24-h time point following a once-off 400 mg oral dose in healthy volunteers also compares well with the amount used in our study. However, the C max of moxifloxacin achieved in epithelial lining fluid was approximately 10 mg/ml (68,92,93), almost double the concentration used in our in vitro model. The plasma C max achieved for moxifloxacin is approximately 4 mg/ml, a value that more closely resembles the value used in this study (94)(95)(96). Bedaquiline, following treatment with a dose of 100 mg, displays plasma concentrations of approximately 0.8 mg/ml, very similar to the 0.6 mg/ml used in this study (97,98). The PhX1 concentration used during this study was very similar to the maximum concentrations observed in both epithelial lining fluid (0.6 mg/ml) and plasma (1.9 mg/ml) obtained in the murine experiments. The similarities of the starting doses of all the compounds to their in vivo concentrations provide a solid grounding for more clinically relevant conclusions to be drawn from these experiments.
Despite moderate to high compound accumulation seen for PhX1 and other compounds, this does not necessarily translate into similar levels of intracellular efficacy (99). This is due to complexities that are increased by cellular involvement, such as drug binding, metabolism, sequestration into organelles, and pH changes. The most notable example of the disconnect between intracellular efficacy and accumulation is reported for moxifloxacin, a drug that shows relatively high accumulation within cells but possesses moderate intracellular activity (86,100). This is again demonstrated in the current study with moxifloxacin displaying moderate efficacy relative to its accumulation, with approximately 0.5 log 10 CFU/ml difference from the starting log 10 CFU/ml in the 96-h assay; this is similar to the case of PhX1, which displayed a 0.8 log 10 CFU/ml decrease. This contrasts with bedaquiline, which showed significant accumulation in infected THP-1 cells and subsequently a log 10 CFU/ml difference of 1.5. Although rifampin did not show the same degree of accumulation, the compound still achieved a bacterial decrease of approximately 1.5 log 10 CFU/ml, highlighting the need to capture both efficacy and accumulation measurements in the same experiments. The PhX1/WHN296 combination showed CFU counts similar to those of the PhX1 alone up until the 48-h time point. Thereafter, the PhX1/WHN296 combination showed significantly more efficacy than the comparator compounds at the remaining 72-and 96-h time points, decreasing bacterial counts by 2 log 10 CFU/ml. This was possibly due to the synergy levels being reached by each compound at the 48-h time point, translating into an increase in efficacy at the same time point. Conversely, synergy levels of the clofazimine/WHN296 combination were reached only by clofazimine and not by WHN296 following the addition of these compounds, resulting in an increase in CFU counts until after the 48-h time point. The redox-active phenothiazine methylene blue, which is structurally related to the phenoxazine PhX1, potently synergizes the action of artemisinins against the malaria parasite through enhanced production of reactive oxygen species (ROS), as described in detail elsewhere (101). Given that M. tuberculosis has a well-defined oxidative stress response (102)(103)(104), it was expected that the redox-active phenoxazine PhX1 or the structurally related clofazimine (34), in combination with the amino-artemisinin WHN296, would promote the production of ROS, thus explaining the synergism seen against M. tuberculosis.
This study supports the use of the combination of PhX1/WHN296, or indeed more lipophilic artemisinin derivatives (29), in further studies given the large degree of both in vitro and in vivo accumulation and efficacy in the in vitro intracellular assays. The model, combining two-dimensional synergy, mass spectrometry, and efficacy evaluation, provides a novel approach to understanding synergistic efficacy and accumulation within the infected macrophage. These combinations may synergistically potentiate intracellular activities of the components offering attractive and more flexible options for combatting clinical TB disease. In addition, it is clear that use of drug combinations capable of enhancing oxidative stress (104) has considerable potential and requires further evaluation according to the methods disclosed here.

MATERIALS AND METHODS
Ethics. All animal studies were conducted with approval from the Animal Ethics Committee of the University of Cape Town (013/032). The experiments were conducted in accordance with the National Code for Animal Use in South Africa (105).
Bronchoalveolar lavage fluid analysis. Animals. Male C57BL/6 mice, 12 to 16 weeks old, weighing approximately 30 g, were acclimatized to their environment 4 days prior to the start of the experiment. Mice were maintained at the animal facility of the University of Cape Town and were fed a standard laboratory diet and water ad libitum. Mice were housed in 27 by 21 by 18 cm cages, under controlled environmental conditions (26 6 1°C with 12-h light/dark cycles).
Dosing. Compounds were prepared in theoretical fixed volumes for the average weight. Dose volumes ranged from 200 to 300 ml for oral dose. Animals (n = 9) were dosed at 10 mg/kg (based on weight), and three animals were dosed with vehicle control (100% hydroxypropyl methylcellulose [HPMC]) in order to obtain blank bronchoalveolar lavage fluid for calibration curves.
Bronchoalveolar lavage fluid procedure. Mice were anesthetized via intraperitoneal (i.p.) injection of ketamine/xylazine (75 to 100 mg/kg plus 10 mg/kg) at predetermined time points (1,8, and 24 h postdrug administration). Depth of anesthesia was monitored by the absence of the pedal withdrawal reflex. Ten min before the bronchoalveolar lavage fluid procedure, blood samples (each 30 ml) were drawn from the mice by tail tip bleeding and placed in tubes containing EDTA. This sample was immediately spun down (5 min, 10,621 Â g) to obtain plasma. Mice were prepared for the surgical procedure by shaving and washing of the neck area. Dissection of the neck tissue to expose the trachea was completed to allow a small tracheal incision allowing the passage of a 23-gauge intravenous (i.v.) catheter into the trachea. PBS (100 ml) was introduced into the lungs of the animals via the lavage tube, aspirated four times inside the lung, and removed. Samples were collected on ice, and the total volume collected was recorded. Exsanguination via cardiac puncture was carried out postprocedure.
Sample processing. Cell numbers were determined using trypan blue staining and a TC-20 automated cell counter (Bio-Rad, CA, USA). Bronchoalveolar lavage fluid samples were then processed by initially spinning down (5 min, 10,621 Â g) the samples to pellet any remaining cells, followed by removal of the supernatant.
ELF volume determination. Bronchoalveolar lavage fluid supernatant was analyzed using the colorimetric Quantichrom urea assay kit DIUR-500 (BioAssay systems, CA, USA). The optical density (OD) of each sample was read using a Bio-Rad iMark microplate absorbance reader (Bio-Rad, CA, USA). A calibration curve was constructed using these readings, and quantitative estimations of urea concentration in both the plasma and bronchoalveolar lavage fluid were determined using this calibration curve. Epithelial lining fluid volume was then determined using equation 1 below.
Determination of epithelial lining fluid volume in murine bronchoalveolar lavage fluid: Sample processing leading to LC-MS/MS analyses. Cellular, plasma, and lung fluid samples were processed by the addition of acetonitrile containing 1 mg/ml of internal standard (carbamazepine) to sample, followed by vortexing (1 min) and centrifugation (5 min, 10,621 Â g). Supernatants were then removed, dried down under nitrogen, and reconstituted for LC-MS/MS analysis using injection solvent.
Standards and QC samples were prepared in each matrix (pooled blank murine matrices) in triplicate and were extracted using the same acetonitrile protein precipitation method as that described above. LC-MS/MS analysis was performed. Briefly, an Agilent 1200 rapid resolution HPLC system comprising a binary pump, a degasser, and an auto-sampler (Agilent, Little Falls, Wilmington, USA) coupled to an AB Sciex 4000 QTrap hybrid triple quadrupole linear ion-trap mass spectrometer (AB Sciex, Framingham, MA, USA) was used for sample analysis. Epithelial lining fluid concentrations were subsequently calculated for each mouse using equation 2 below.
Concentration of drug in epithelial lining fluid (ELF): Uninfected macrophage assay. Cell culture and drug addition. Tamm-Horsfall protein 1 cells (THP-1) were plated in 24-well plates, followed by the addition of 0.1 mM phorbol 12-myristate 13-acetate (PMA). The cells were then supplemented with RPMI medium containing 10% fetal calf serum (FCS). Cells were grown at 37°C in 5% CO 2 . RPMI containing 20Â M. tuberculosis MIC 90 of the drugs (compound-specific) was added in a volume of 500 ml, in triplicate. This was followed by washing of cells using RPMI medium and was repeated thrice. EDTA was used to detach cells from the surface of the plate. Cell numbers were determined using a trypan blue staining method and a TC-20 automated cell counter (Bio-Rad, CA, USA). Cells were then lysed using 100 ml chloroform/methanol (3:1, vol/vol) and 1 mg/ml of compound-specific internal standard (deuterated internal standard for known anti-TB compounds and carbamazepine for novel compounds).
LC-MS/MS sample preparation and analysis. Samples were vortexed (1 min) and centrifuged (5 min, 10,621 Â g). Sample supernatants were subsequently removed, transferred into a 96-well analysis plate, and dried down under nitrogen. Samples were reconstituted in a mixture of 1:1 H 2 O/acetonitrile (containing 0.1% formic acid [FA], vol/vol) and processed using quantitative LC-MS/MS assays developed at the University of Cape Town. Chromatographic separation was performed using an Agilent 1200 rapid resolution HPLC system consisting of a binary pump, a degasser, and an auto-sampler (Agilent, Little Falls, Wilmington, USA) coupled to an AB Sciex 4000 QTrap, hybrid triple quadrupole linear ion-trap mass spectrometer (AB Sciex, Johannesburg, South Africa). Blank cell lysates (lysed using acetonitrile [0.1% Triton X]) were spiked in triplicate with serially diluted compound concentrations to generate the calibration (2 to 5,000 ng/ml) and quality control (6 to 4,000 ng/ml) samples. These samples were extracted using the procedure described above and were used to construct calibration curves to quantitate the amount of analyte in each respective intracellular sample. Extracellular concentrations were determined by spiking serially diluted concentrations of compound into media removed from the untreated cells to generate calibration and quality control samples and used to determine drug concentrations in the extracellular environment.  (Tables S1 to S3).
In vitro cytotoxicity in THP-1 cell lines. THP-1 cells were routinely maintained as adherent monolayers as reported above. The culture medium was changed every 3 days when the cells were subcultured. Mosmann's MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide] assay was used, with minor modifications to determine cell viability (113). After the addition of the compounds at a starting concentration of 50 mM (serial diluted to 5 mM), each well received MTT at a concentration of 5 mg/ ml in phosphate-buffered saline (PBS), with blank samples receiving only medium and MTT. Each compound concentration tested in this study was completed in triplicate. The plates were incubated for 4 h at 37°C before centrifuging for 10 min at 500 Â g and removing the supernatant from the wells without disturbing the formazan crystals. DMSO was added to each well, and the plate was shaken for 5 min on vortex to dissolve the crystals. The absorbance of the formazan salt was measured at 540 nm by a Bio-Rad iMark microplate absorbance reader (CA, USA). The following formula (equation 3) was used to calculate the cell viability: Cell viability in CHO and VERO cell lines: % viability ¼ sample absorbance 2 blank absorbance control absorbance 2 blank absorbance Â 100 Nonlinear dose response curves were constructed using GraphPad Prism 4 software and Microsoft Excel. Fluorescence microscopy. THP-1 cells were seeded in a 24-well plate containing a single pretreated (poly-L-lysine) cover slip at 5 Â 10 5 cells/well. After PMA maturation for 24 h, the cells were infected with a log-phase culture of green fluorescent protein (GFP)-expressing M. smegmatis (OD = 0.6) at a multiplicity of infection (MOI) of 5. The infection was allowed to continue for 30 min, after which the wells were washed with fresh RPMI medium. RPMI medium containing 1 mg/ml of each drug was added to each well, in triplicate. The experiment progressed for an additional 2 h before the wells were again washed with RPMI medium. The coverslips were carefully removed from each well and adhered to a microscope slide containing a single drop of paraformaldehyde. Microscope slides were then examined using an Axio Scope A1 microscope and images were captured using a Zeiss 1 MP monochrome camera (Carl Zeiss, Germany). Samples were exposed to the same exposure settings for respective channels for image acquisition. GFP fluorescence was detected using a 488-nm excitation filter and a 510-nm emission filter set. Red fluorescence from the compounds was detected using a 587-nm excitation laser filter and 610nm long-pass emission filter. Finally, brightfield imaging was used to detect the position of the bacteria within the macrophage cells.
THP-1 cells were seeded at a concentration of 5 Â 10 5 cells/well. After PMA maturation for 24 h, the cells were infected with a log-phase culture of Mycobacterium smegmatis mc2155::gfp (OD = 0.6), a mycobacterial reporter strain expressing GFP protein (114,115) at a multiplicity of infection (MOI) of 5. The infection was allowed to continue for 30 min, after which the wells were washed with fresh RPMI medium. RPMI medium containing 1 mg/ml of drug was added to each well. The experiment was allowed to progress for an additional 2 h before the wells were again washed with RPMI medium. Cells were detached using EDTA (5 mM). Samples were left unfixed, as fixation agents may have caused cell lysis and loss of drug. Sorting was completed using the FACS Jazz cell sorter instrument for 10,000 events per samples (BD, NJ, USA). Gating was initially based on forward and side scatter (size of macrophage) to differentiate debris and undifferentiated cells, followed by further gating based on the live/ dead stain SYTOX (used according to the manufacturer's instructions). Samples were analyzed and further sorted based on GFP fluorescence, providing both uninfected and infected populations of cells. Results were analyzed using the FlowJo software package.
M. tuberculosis-infected macrophage assay. THP-1 cells were grown in RPMI medium containing 10% FCS at 37°C in 5% CO 2 . Cells were seeded and differentiated according to the above protocol for 24 h using PMA. M. tuberculosis (H37RvMa) cells were grown to mid-log phase in 7H9 medium. THP-1 cells were infected at an MOI of 5. Following 3 h of infection, wells were washed with RPMI medium to remove extracellular bacteria and dead cells. RPMI medium containing a predetermined concentration of compound was added to each well in triplicate. The solution was allowed to equilibrate for 30 min. In addition, three wells were left without drug treatment (following M. tuberculosis infection) and served as untreated controls, while an additional three wells of THP-1 cells were left uninfected and were used as uninfected controls. Thereafter, at each time point, cells were washed with RPMI medium. This was followed by the addition of EDTA (5 mM), after which 5 min was allowed for the cells to detach. Trypan blue-stained cells were counted using the automated cell counter. The remaining supernatant (150 ml) was added directly to an Eppendorf tube containing chloroform/methanol (3:1, vol/vol) and 1 mg/ml of compound-specific internal standard (deuterated internal standard for known anti-TB compounds and carbamazepine for novel compounds). An aliquot of 5 ml of the remaining supernatant was added to 90 ml RPMI medium, and 10 ml of this solution was streaked onto hygromycin-containing 7H10 plates and placed at 37°C, followed by counting 21 days following plating. After the collection of samples at the remaining time points, the samples were surface-decontaminated and stored at 280°C overnight. The samples were thawed and transferred into a 96-well analysis plate and dried down under a gentle stream of NO 2 gas. The samples were then reconstituted in a mixture of 1:1 acetonitrile/H 2 O (containing 0.1% FA, vol/vol) before analysis via LC-MS/MS. Samples were processed using quantitative LC-MS/MS assays. Chromatographic separation was performed using an Agilent 1200 rapid resolution HPLC system consisting of a binary pump, degasser, and auto-sampler (Agilent, Little Falls, Wilmington, USA) coupled to an AB Sciex 4000 QTrap, hybrid triple quadrupole linear ion-trap mass spectrometer (AB Sciex, Johannesburg, South Africa). Blank infected cell lysates (lysed using chloroform:methanol) were spiked in triplicate and extracted using the same procedure as that described above to generate compoundspecific standard and quality control samples. These were then used to construct calibration curves, which were subsequently used to quantitate the amount of analyte in each respective sample. Extracellular concentrations were determined by spiking serially diluted concentrations of compound in triplicate into media removed from untreated THP-1 cells to generate calibration and quality control samples, which were then extracted using the same procedure as that described above and used to quantitate the amount of compound in each extracellular sample.
Synergy assay. Drug interactions were assessed by checkerboard dilution in a 96-well format with slight modifications to the method described by Chen et al. (116). The first column contained drug-free control cells and the last column contained control drug giving maximum inhibition. Compound A was pipetted at starting concentrations 100-fold higher than those in the 96-well plate along the x axis (column 3 to 11), and compound B was pipetted at starting concentrations 50-fold higher than those in the 96-well plate (from row B to H). Column 1 and column 12 contained maximum inhibitory concentrations of drug (no bacterial growth) and media containing only M. tuberculosis (maximum bacterial growth), respectively. Rows A3 to A11 contained only PhX1, and columns B2 to H2 contained only WHN296, which allowed for the calculation of MIC 90 values for each compound. M. tuberculosis cultures were overlaid (100 ml) onto the drug containing wells at a concentration of 1 Â 10 8 cells/ml (OD = 0.6). Plates were incubated at 37°C for 8 days. Following the addition of resazurin, the plates were incubated for a further 3 days to allow a color change from blue to pink, associated with mycobacterial growth, and read at 560 nm on a Spectramax multimode plate reader (Molecular Devices). Synergy was determined by the fractional inhibitory concentration index (FICI) in equation 4. FICI ratio determination: where CAcomb and CBcomb are the concentrations of drugs A and B which are the iso-effective concentrations (combinations of each drug which produced a level greater than their respective MIC 90 ). FICI values for which the FICI was #0.5 represented synergy, FICI of .4 represented antagonism, and any value between 0.5 and 4 represented no interaction (117).

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 0.5 MB.

ACKNOWLEDGMENTS
We acknowledge the following groups and individuals for their contributions to this study: the H3D ADME and PK teams and Trevor We declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.