Subcellular Partitioning and Intramacrophage Selectivity of Antimicrobial Compounds against Mycobacterium tuberculosis

ABSTRACT

The efficacy of antimicrobial drugs against Mycobacterium tuberculosis, an intracellular bacterial pathogen, is generally first established by testing compounds against bacteria in axenic culture. However, inside infected macrophages, bacteria encounter an environment which differs substantially from broth culture and are subject to important host-dependent pharmacokinetic phenomena which modulate drug activity. Here, we describe how pH-dependent partitioning drives asymmetric antimicrobial drug distribution in M. tuberculosis-infected macrophages. Specifically, weak bases with moderate activity against M. tuberculosis (fluoxetine, sertraline, and dibucaine) were shown to accumulate intracellularly due to differential permeability and relative abundance of their ionized and nonionized forms. Nonprotonatable analogs of the test compounds did not show this effect. Neutralization of acidic organelles directly with ammonium chloride or indirectly with bafilomycin A1 partially abrogated the growth restriction of these drugs. Using high-performance liquid chromatography, we quantified the degree of accumulation and reversibility upon acidic compartment neutralization in macrophages and observed that accumulation was greater in infected than in uninfected macrophages. We further demonstrate that the efficacy of a clinically used compound, clofazimine, is augmented by pH-based partitioning in a macrophage infection model. Because the parameters which govern this effect are well understood and are amenable to chemical modification, this knowledge may enable the rational development of more effective antibiotics against tuberculosis.

INTRODUCTION

Mycobacterium tuberculosis, the causative agent of tuberculosis (TB) and HIV, are the two pathogens which are the most prolific killers of humans (1). As many as two billion people are estimated to be latently infected with M. tuberculosis (2), despite the existence of effective antibiotics. Treatment of TB is challenging due to the long duration of chemotherapy needed to sterilize infection (3) and the emergence and spread of drug resistance (4). The current standard regimen for uncomplicated, drug-susceptible cases was shortened from 9 to 12 months to the current 6-month short-course period with the addition of pyrazinamide (PZA) (3, 5). Strikingly, PZA is inactive in vitro at the neutral pH used in most whole-cell screens in culture medium and so would not have been identified by current approaches. It is only because PZA was administered in a murine model of TB that its potent in vivo activity was discovered (5). This demonstrates the need for more robust systems for screening potential novel antimicrobial drugs in relevant pharmacokinetic environments. Screening compounds against M. tuberculosis in its natural niche of the macrophage phagosome represents progress to that end, and this study derives from a hit from such an intramacrophage screen (6).

Specifically, the antidepressant drug fluoxetine (FLX) was found to restrict M. tuberculosis growth inside macrophages at concentrations much lower than those required for direct antibacterial effect against bacteria in culture medium. While this effect was originally thought to indicate a host-mediated effect, several experimental approaches suggest that direct action of the compound occurs due to accumulation inside the acidic compartment of exposed cells.

Because weakly basic compounds of moderate lipophilicity have been shown to accumulate in acidic organelles through a phenomenon called ion trapping (7–10) and because M. tuberculosis is known to reside in a relatively acidic compartment (11, 12), we investigated whether this mechanism explains the increased activity of FLX and other chemically similar compounds when applied to infected macrophages relative to their activity against bacteria in broth culture. In order to make conditions as comparable as possible, we adjusted the pH of the nutrient broth slightly to match the approximate value encountered by M. tuberculosis inside phagosomes, i.e., pH 6.4 (13, 14). We will refer to this class of compounds as ACA, for acidic compartment accumulating drugs. They are characterized by pK_a values which are basic relative to the acidic compartment (generally 6.5 to 11) and CLogP values above approximately 2. Beyond FLX, two additional drugs were examined based on their pK_a and CLogP values, which are consistent with an ACA profile. They are the local anesthetic dibucaine (DIB) and the antidepressant sertraline (SRT) (15, 16). See Table 1 for their chemical properties and Fig. 1 for their structures.

TABLE 1.

Growth-inhibitory activity of study drugs in broth medium and inside macrophages

Compound	pKa	CLogP	LogD at pH:		Predicted ACA profile?	Broth IC50 (μM)	MIC (μM)	IM-IC50a (μM)	Selectivity indexb
			7.4	6.4
SRT	9.7	5.2	3.0	2.3	Yes	137	59.0	9.8	6.0–14
kSRT	<0	4.9	4.9	4.9	No	355	125	Inactive	NA
FLX	9.8	4.2	1.8	1.2	Yes	164	69.4	7.9	8.8–21
qFLX	<0	0.4	0.4	0.4	No	305	194	Inactive	NA
DIB	9.0	3.7	2.1	1.1	Yes	771	104	10	10–77
INH	3.4	−0.69	−0.69	−0.69	No	0.31	0.29	0.63	0.46–0.49

^aIM-IC₅₀ stands for intramacrophage IC₅₀. Inactive indicates no statistically significant growth inhibition at the highest concentration tested (40 μM).

^bSelectivity index was calculated as broth IC₅₀/IM-IC₅₀ and as MIC/IM-IC₅₀. NA, not applicable.

FIG 1.

(A) Structures of study compounds. Intramacrophage growth restriction of M. tuberculosis by luminescent reporter (n ≥ 3; error bars indicate ± standard errors of the means [SEM]) (B) or CFU enumeration (n = 2; ±SEM) (C). All treatments show reduced relative luminescent units (RLU) or CFU compared to the DMSO control (P < 0.05 by analysis of variance [ANOVA] with Bonferroni's adjustment for multiple comparisons). The highest dose (10 μM) of each drug is more efficacious than the lowest dose (2.5 μM) by unpaired t test within each group (P < 0.05). Study compounds do not induce cell death by CellTox Green assay (D) (n = 2 to 4; ±SEM; all comparisons versus the DMSO control were nonsignificant except for the Triton X-100 positive control for membrane integrity loss, which had a P value of <0.001 [***]). All conditions in each experiment were tested as triplicate biological replicates; CFU experiments were plated in triplicate from triplicate biological replicates.

RESULTS

M. tuberculosis growth restriction inside macrophages by ACA compounds.

All three compounds with predicted accumulation in acidic organelles (Fig. 1A) were found to restrict the growth of M. tuberculosis after 72 h of treatment in resting primary murine macrophages in a dose-dependent manner based on a bioluminescent reporter strain (Fig. 1B). The level of restriction observed with FLX was comparable to the original description of its activity in this context (6) and with results achieved by CFU enumeration (Fig. 1C). For all ACA compounds, the highest dose was significantly more efficacious than the lowest dose, indicating a dose-response gradient.

Activity of ACA drugs is not due to host cell toxicity.

In order to rule out the possibility that the observed effect was secondary to a cell death phenotype, we tested for macrophage cell death using the fluorescent dye CellTox green, which is excluded from living cells but is highly fluorescent when bound to the DNA of dead or dying cells. No significant cell death was observed relative to solvent control under ACA drug treatment conditions, while cells exposed to the detergent Triton X-100 had a strong fluorescent signal, indicative of loss of plasma membrane integrity (Fig. 1D).

Neutralization of acidic compartments reduces growth-inhibitory effect of ACA compounds.

To determine whether neutralization of the pH gradient would reduce or eliminate accumulation of ACA drugs, we used ammonium chloride to directly neutralize acidic compartments of infected, resting macrophages just after infection (17). Separately, we used bafilomycin A1 to achieve the same effect indirectly through inhibition of the vacuolar H⁺ ATPase (18). We titrated the concentrations such that the neutralization treatments alone did not affect bacterial load as measured by luminescence assay. Statistically significant reductions in the restriction of mycobacterial growth were observed with ACA compounds when they were applied with ammonium chloride (20 mM) or bafilomycin A1 (5 nM), while a clinical antimicrobial drug (isoniazid [INH]), which has a chemical profile not consistent with pH-based partitioning, was not affected by neutralization (Fig. 2).

FIG 2.

Neutralization of acidic organelles either directly with ammonium chloride or indirectly with bafilomycin A1 partially abrogates the growth restriction observed with FLX, SRT, and DIB (10 μM). The effect of low-dose (0.15 μM) isoniazid (INH) is not reduced by neutralization with NH₄Cl. (*, P < 0.05; **, P < 0.005; ****, P < 0.0001; all by ANOVA with Bonferroni adjustment for multiple comparisons; ns, not significant by unpaired, two-tailed Student's t test).

Nonprotonatable analogs of ACA compounds are inactive against intramacrophage bacteria.

To confirm that the secondary amines in FLX and SRT are the basic functional group governing the growth restriction observed, we tested the activity of close analogs lacking that group. Specifically, we tested an analog of FLX which has a quaternary amine rather than a secondary amine and is therefore positively charged at physiologic pH (referred to as qFLX) and an analog of sertraline which has a ketone in place of the secondary amine (referred to as kSRT) (Fig. 3). These compounds showed comparable extracellular growth-inhibitory activity to the parent compounds (not more than 3-fold higher values for broth 50% inhibitory concentration [IC₅₀] or MIC) but showed no statistically significant restriction of mycobacterial growth inside primary murine bone marrow-derived macrophages (Fig. 3 and Table 1). When applied to infected macrophages, the parent compounds were still active at 16-fold lower concentration (2.5 μM) than the highest tested analog concentrations (40 μM) which failed to restrict growth of M. tuberculosis. These data suggest that the slight decrease in extracellular activity of the analogs does not account for their absent intramacrophage activity.

FIG 3.

Nonprotonatable analogs of study drugs (A) do not restrict mycobacterial growth (B) by luminescence assay for M. tuberculosis bacterial load (n ≥ 3; ±SEM). ****, P value of <0.001 by ANOVA with Bonferroni's adjustment for multiple comparisons. Stepwise comparisons within each analog group are nonsignificant (P value of >0.5 by the same test).

An M. tuberculosis mutant which resides in a more acidic intracellular environment is more sensitive to FLX and SRT within macrophages despite being no more sensitive when extracellular.

If accumulation of compounds in this study by partitioning into acidic compartments is the mechanism of M. tuberculosis growth restriction, it should follow that M. tuberculosis mutants with greater exposure to highly acidic environments inside macrophages would be more sensitive to them. Since such mutants have been described, we tested one which has been reported to reside in a more mature phagolysosome-like intracellular environment (pH ∼5.8; pH ∼6.4 for the wild type) without profound attenuation (13). This mutant (for this study, produced by transposon insertion into Rv2930c and referred to as Rv2930c::Tn) was found to have an indistinguishable IC₅₀ by optical density growth curve as well as MIC by microtiter broth dilution to strain-matched wild-type M. tuberculosis H37Rv. The mutant was, however, more sensitive to both FLX and SRT when applied to infected macrophages (Fig. 4). Dibucaine was not more effective versus Rv2930c::Tn. All drugs had similar or reduced activity relative to the control at pH 6.0 versus 6.5 or 7 in growth curve experiments, which precludes the possibility of the increased restriction being the result of pH dependence of drug activity (data not shown).

FIG 4.

Transposon mutant of M. tuberculosis, Rv2930c::Tn (empty bars), which resides in a more acidic compartment within macrophages, is more sensitive to FLX and SRT than the wild-type control (H37Rv, filled bars). P values were 0.004 (**) and <0.001 (****) by unpaired, two-tailed t test with Welch's correction.

Clofazimine, an antimicrobial drug used in TB treatment, is subject to ion trapping.

The chemical properties of clofazimine, a second-line TB drug (19), are consistent with an ACA profile (Fig. 5A and B), so we investigated whether the compound's activity was augmented by pH-based partitioning. We established that a concentration of 0.5 μg/ml reduces bacterial load by approximately 46% relative to solvent control in infected macrophages. We found that this reduction was partially reversible with ammonium chloride or bafilomycin A1 treatment, to 65% and 78% of the control, respectively (Fig. 5C). None of the treatments caused significant cell death by CellTox green assay; a detergent control (Triton X-100; 0.1% final concentration) was the only treatment statistically different from solvent control (Fig. 5D). Clofazimine has a fluorescence spectrum which allowed us to image the accumulation of the drug inside presumed acidic compartments of the cell. Using wide-field epifluorescence microscopy (excitation at 550 nm, emission at 605 nm), we visualized resting primary murine macrophages after overnight incubation with clofazimine alone (5 μg/ml) or in combination with ammonium chloride (20 mM) or bafilomycin A1 (5 nM) (Fig. 6A). While a weak autofluorescence signal could be observed in the solvent control and some amorphous staining was observed under the neutralization conditions, the clofazimine-alone condition showed a much stronger overall fluorescence signal and a punctate staining appearance consistent with lysosomal accumulation (Fig. 6A). Using a stringent threshold applied via a macro for ImageJ to objectively quantify puncta above a set brightness level, the clofazimine-alone condition was found to contain significantly more such areas per cell than under those conditions where bafilomycin or ammonium chloride was added (Fig. 6B).

FIG 5.

M. tuberculosis growth restriction by clofazimine (the structure is shown in panel A, with protonatable nitrogen indicated by boldface, and physiochemical properties are listed in panel B) is partially reversed by acidic compartment neutralization in infected BMDM by luminescence assay for bacterial load (C) (n ≥ 3; ±SEM) without altering macrophage viability as determined by CytoTox Green assay (D). The Triton X-100 (TX-100) positive control for membrane integrity loss shows significant increase in fluorescent signal. **, P value of <0.01 by ANOVA with Bonferroni's adjustment for multiple comparisons.

FIG 6.

Overnight incubation induces observable accumulation of clofazimine in BMDM. (A) Clofazimine accumulation is partially reversible by acidic compartment neutralization with NH₄Cl or bafilomycin A1. Excitation and emission wavelengths, 550 and 605 nm, respectively. (B) Quantitation of puncta in >3,300 cells, including at least 700 from each treatment, using ImageJ software. **, P value of <0.01 by ANOVA with Bonferroni's adjustment for multiple comparisons.

Protonatable ACA compound accumulation in macrophages is partially pH dependent.

Tissue-level accumulation, especially in lung and spleen, has been demonstrated for established (20, 21) and emerging (22, 23) antitubercular compounds and is believed to enhance their efficacy against M. tuberculosis. We sought to determine if ACA compounds accumulate within bone marrow-derived macrophages (BMDM), and if so, the extent to which such accumulation depends on acidic compartments. Resting, uninfected, or infected BMDMs were exposed to SRT, kSTR, FLX, qFLX, or clofazimine (10 μM) for 24 h under nonneutralizing or neutralizing (20 mM NH₄Cl) conditions. Analytes were extracted from macrophages, and the cellular concentration/medium concentration ratio was compared to that of the vehicle control (Fig. 7). SRT, FLX, and clofazimine all showed accumulation that was stable enough to persist during the analyte extraction workup. Collecting cells by scraping at 4°C, rather than trypsin treatment at 37°C, did not affect the measured accumulation of a test compound (clofazimine; data not shown). Accumulation was partially reversed by neutralizing acidic compartments. The partial attenuation of accumulation under neutralizing conditions suggests that SRT, FLX, and clofazimine accumulate in acidic compartments but that acidic compartments are not the only reservoir for these compounds within cells. Nonprotonatable analogs kSRT and qFLX did not accumulate stably under either condition. Infected macrophages showed significantly greater accumulation of SRT, FLX, and clofazimine than uninfected cells, and that accumulation was similarly attenuated by neutralization with ammonium chloride.

FIG 7.

ACA compounds accumulate in macrophages (C_c, cellular concentration) relative to medium (C_m, medium concentration; 10 μM). BMDM were exposed for 24 h to sertraline (SRT), sertraline ketone (kSRT), fluoxetine (FLX), quaternary fluoxetine (qFLX), or clofazimine (Clof) with or without ammonium chloride (20 mM) after 4 h of infection with M. tuberculosis (multiplicity of infection, 1:1) or mock infection. Analytes were quantified by UV absorbance (200 nm) during LC-MS analysis. SRT, FLX, and Clof accumulation was partially reversible by ammonium chloride treatment, and accumulation was greater in M. tuberculosis-infected cells (**, P value of <0.01 by ANOVA with Bonferroni's adjustment for multiple comparisons). Ammonium chloride also decreased accumulation in infected cells relative to infected cells without neutralization treatment. ***, P value of <0.001 by unpaired two-tailed t test with Welch's correction (n ≥ 3; ±SEM).

DISCUSSION

In his seminal 1974 work on the subject, Christian de Duve proposed that antibiotics with “lysosomotropic” properties are more effective against intracellular pathogens (9). Despite the passage of 4 decades, this is, to our knowledge, the first investigation of that theory with regard to mycobacterial infection.

Weak bases are more membrane permeable in their nonionized form (24) but become trapped inside acidic compartments when protonated. Actively maintained pH gradients give rise to orders-of-magnitude-sized asymmetric distributions of weak bases across cellular membranes such as lysosomal membranes. Dibucaine, a compound used in this study to restrict M. tuberculosis growth in macrophages, exhibits this behavior in agreement with de Duve's theory in large unilamellar vesicles, which are a model for membrane-enclosed compartments (15).

Although other groups have looked at the impact of pH-based partitioning on antibiotic pharmacokinetics generally (25, 26) and apparent accumulation of drugs used in the treatment of TB has been observed (27, 28), the basis for this accumulation has never been demonstrated in the context of mycobacterial infection.

Our results demonstrate intracellular accumulation of several drugs to approximately 200 to 400 times the concentration present in the extracellular space. This accumulation was partially dependent on the pH gradient (Fig. 7). This degree of concentration enables weakly antimycobacterial compounds fluoxetine, sertraline, and dibucaine to restrict M. tuberculosis in macrophages at concentrations well below the level needed for these same compounds to restrict M. tuberculosis in broth cultures (Fig. 1 and Table 1). Nonprotonatable analogs of these compounds were only able to restrict M. tuberculosis in broth and were not seen to accumulate in macrophages, although we cannot rule out the possibility that dissipation of non-ACA compounds into wash buffers during workup obscured rapidly reversible accumulation. The enhanced accumulation (200- to 400-fold) of ACA compounds in M. tuberculosis-infected versus uninfected cells suggests this phenomenon influences TB drug pharmacokinetics generally despite the well-documented ability of M. tuberculosis to limit acidification of its phagosomal environment.

To investigate whether this effect governs the selectivity of clinical TB drugs, we tested the predicted ACA compound clofazimine and the predicted non-ACA compound isoniazid. Clofazimine has been shown to be more active against intracellular bacteria in the context of Mycobacterium avium infection (29), and here we demonstrate pH-dependent restriction (Fig. 5), accumulation (Fig. 7), and localization into large puncta (Fig. 6) with that compound in M. tuberculosis-infected murine macrophages. However, we were unable to demonstrate colocalization of clofazimine with the acidic compartment due to spectral overlap of the LysoTracker dyes and the loss of subcellular localization upon aldehyde fixation and permeabilization, which precluded immunostaining. Without colocalization images, the demonstration of acidic compartment localization is inferred from the functional data rather than direct visualization. The non-ACA compound isoniazid has been shown to be present at similar or lower concentrations inside macrophages relative to the extracellular space (27), and consistent with that finding, our results show no pH dependence in isoniazid restriction of M. tuberculosis during macrophage infection (Fig. 2).

The partial neutralization of (phago)lysosomes by ACA drugs due to their basicity may represent a countervailing effect to the direct antimicrobial action of the drugs themselves (12) and could account for the lack of additional effect against the Rv2930c::Tn mutant demonstrated by dibucaine in this study (given its relatively low antimicrobial activity but high accumulation). This effect, if present, would decrease with more potently antimicrobial ACA drugs, though, since they would require much lower molar concentrations and would therefore have less alkalinizing effect than the ACA drugs in this study. It has been shown that chloroquine, a known lysosomal accumulator, does not decrease the activity of pyrazinamide, a drug only active at acidic pH, when applied to infected macrophages (30). This suggests that strong alkalization of M. tuberculosis-containing structures does not occur with that compound and may be generalizable. That finding, and our results with ammonium chloride and bafilomycin treatment alone, which were similar to published results reported by another group (18), suggests that acidic compartment neutralization alone does not profoundly impact the course of infection of macrophages with wild-type M. tuberculosis. Thus, while mutant M. tuberculosis strains which traffic to more mature phagolysosomes may be rescued by neutralization (18), wild-type bacteria are less affected. Still, our results with bafilomycin showed enough variability that we felt it was necessary to pursue a second neutralization mechanism to confirm the result (e.g., Fig. 2 results for Baf in combination with SRT and DIB).

Our results do not explain completely the mechanism by which ACA compounds achieve enhanced restriction. Accumulation of drugs to levels that are directly bactericidal is only one possible mode of action. Physical accumulation is, however, well supported by observations regarding clofazimine in the clinical literature (31–34), wherein crystals of drug have been observed in lung tissue of treated patients. Interest in clofazimine has recently intensified due to its successful use against multidrug-resistant M. tuberculosis (35, 36) and the demonstration that it can shorten treatment in a murine model of TB (37). Still, an effect on host cells cannot be totally excluded, and evidence exists to suggest that loading lysosomes with clofazimine impacts the expression of lysosomal enzymes (38).

The impact of ACA drugs on the endogenous functions of the structures they accumulate within is difficult to predict. Host-directed effects that are beneficial for restriction may also result from ACA drug deposition such as phopholipidosis (39), or, more speculatively, expanded lysosomal volume, resulting in interference with the close apposition of the bacterial cell wall to the host membrane, which has been reported to be necessary for phagosomal maturation arrest (40), a key survival strategy of M. tuberculosis.

Other potentially relevant host side effects could include an impact on autophagy, in particular, given that bafilomycin is an inhibitor of that process and was shown to partially reverse the action of ACA drugs. A major role for canonical autophagy is essentially ruled out, however, by our observations of unaffected activity of FLX and SRT in autophagy-deficient (ATG5 conditional knockout) murine macrophages (see Fig. S1 in the supplemental material).

Specific host-targeted effects of ACA drugs must also be considered. The antidepressant FLX is thought to act by blocking serotonin uptake from synapses (41). Although murine macrophages may express the serotonin transporter (42), the level of serotonin in culture medium did not impact the mycobacterial growth we observed in our infections at any physiologic concentration, and medium made with dialyzed, serotonin-free fetal bovine serum (FBS) performed no differently from standard FBS (Fig. S2). Also, the concentration range used in this study was around three orders of magnitude greater than the K_i of the drug for the serotonin transporter (43). At that level, FLX should saturate practically all binding sites, and a dose-wise response should not be observed. It remained possible that FLX was directly binding serotonin receptors, rather than the transporter, because such affinity has been reported (43). However, the enantiomers of FLX, which have differing affinity for many of the serotonin receptors from one another, have the same ability to restrict the growth of M. tuberculosis in macrophages (Fig. S3). These data suggest that specific binding of FLX to host targets is not the mechanism by which it restricts mycobacterial growth.

A multifactorial mechanism of restriction of mycobacterial growth by high-accumulating drugs seems likely, with some factors favoring restriction and others subverting restriction depending on the compound. Our contention here is not that FLX, SRT, and DIB are viable drugs for the treatment of TB. Indeed, they do not appear to be clinically relevant despite previous reports of their antimicrobial activity (44–47). Only for cryptococcal infection has the therapeutic index of one of the drugs (i.e., SRT) been suggested to be compatible with clinical treatment (47). We used them to demonstrate a phenomenon which could increase the local concentration of antimicrobial drugs to which M. tuberculosis is exposed intracellularly.

Rational modification of antibiotics to endow them with an ACA profile has been demonstrated; Renard and colleagues modified penicillin G (an otherwise nonaccumulating, predominantly extracellular drug) by removing a carboxylic acid moiety and replacing it with a basic, amine functional group. That alteration was sufficient to cause a shift toward accumulation of the drug in the acidic compartment, primarily lysosomes, as identified by cofractionation (48).

Drugs tailored to fit the ACA profile would be expected to be more effective than unmodified versions against intracellular M. tuberculosis. However, accumulation alone does not guarantee increased activity, since differences in the site of accumulation may lead to divergent pharmacodynamic profiles. This could be due to macromolecular binding, especially to membranes given that ACA compounds are by definition somewhat hydrophobic, or due to other factors, such as the ion content of the new site of accumulation, including pH differences. Bearing these caveats in mind, application of the principles described in this study may aid the rational development of antimicrobial drugs for mycobacterial infection.

MATERIALS AND METHODS

Macrophage cell culture.

All experiments were performed on resting primary murine bone marrow-derived macrophages (BMDM) from C57BL/6 mice at a density of 5 × 10⁴/well in 200 μl of medium in 96-well cluster plates. BMDM medium was composed of Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (references 11960 and 10082; Gibco), glutamine as GlutaMAX (reference 35050; Gibco) at 2 mM, and 10% 3T3-macrophage colony-stimulating factor cell conditioned medium. Macrophages were maintained at 37°C, 5% CO₂ in a humidified incubator, and handling steps were executed as quickly as safely possible to minimize exposure to ambient conditions. For cytotoxicity measurement, CellTox green was used according to the manufacturer's instructions (G8741; Promega).

Bacterial culture.

All experiments were conducted with M. tuberculosis grown to mid-log phase. Bacteria were grown in 50 ml of Middlebrook 7H9 broth (271310; Difco) supplemented with 0.5% (vol/vol) glycerol, 0.05% (wt/vol) Tween 80 (P1754; Sigma), and 10% (vol/vol) oleate-albumin-dextrose-catalase enrichment (Middlebrook OADC; catalog no. 212351; BBL). Cultures were maintained at 37°C in 1-liter roller bottles. The Erdman strain of M. tuberculosis was used for all experiments except for those performed for liquid chromatography-mass spectrometry (LC-MS) analysis and for those involving the Rv2930c::Tn mutant, for which the strain-matched control was H37Rv. The autoluminescent strain of M. tuberculosis Erdman, carrying the luxCDABE operon, was provided by Jeffrey Cox. The Rv2930c::Tn mutant was a generous gift of Sarah Stanley.

MIC experiments and growth curves.

MICs were determined by a standard microtiter dilution protocol (49). Extracellular IC₅₀s were determined by monitoring growth of M. tuberculosis using optical density (600 nm) in 240 μl of complete 7H9 in 96-well plates in triplicate across a range of concentrations for each study drug and extrapolating the value which gave half-maximal inhibition. For growth curves, plates were maintained at 37°C with 60-rpm circular agitation. All MIC experiments used 7H9 medium at pH 6.4 to mimic the physiologic pH of the niche inhabited by M. tuberculosis.

Infection.

Bacteria were washed thrice in phosphate-buffered saline (PBS), resuspended, and centrifuged at 50 × g to pellet clumps. Bacterial suspension was added to DMEM with 10% horse serum substituted for FBS to produce infection medium at 5 × 10⁵ CFU/ml in 100 μl per well. Phagocytosis was allowed to proceed for 4 h, after which macrophages were washed twice with warm PBS and returned to BMDM growth medium.

Drug treatment.

Immediately after infection, macrophages were returned to BMDM medium with or without experimental compounds as appropriate. For experiments involving neutralization of acidic organelles, macrophages were returned to 100 μl of BMDM medium with or without neutralization agent for 1 h, followed by a further 100 μl of medium containing 2-fold concentrated experimental compound or control dissolved in BMDM medium of the same composition as the first 100 μl to yield the target concentration of experimental drug.

Bacterial load quantification.

To enumerate bacterial CFU, we lysed macrophages by adding 50 μl of 0.5% Triton X-100 to give a final concentration of 0.1% detergent. Serial dilutions were produced in complete 7H9 medium and plated on Middlebrook 7H10 agar plates (262710; Difco) containing 10% OADC and no antibiotics. CFU were counted after 17 to 21 days of incubation at 37°C in a humidified incubator. For luminescence experiments, macrophages were maintained in white 96-well cluster plates (3610; Costar) with white vinyl sealing tape attached to the bottom (236272; Thermo Scientific) and read at 32°C with a SpectraMax L luminometer (Molecular Devices). Triplicate reads were averaged for each of three biological replicates at each time point.

Reagents.

The following compounds were purchased from Sigma-Aldrich: dibucaine (D0638), resazurin (R7017), ammonium chloride (A-9434), bafilomycin A1 (B1793), isoniazid (I3377), racemic fluoxetine (F132), serotonin (H9523), Triton X-100 (T8787), and clofazimine (C8895). Fluoxetine enantiomers and the quaternary amine analog were generous gifts of Eli Lilly and Co. Sertraline was from Enzo (BML-NS115), and the ketone analog was from Matrix Scientific (072144). Hoechst 33342 was from Anaspec (AS-83218). Calcein AM (C3100MP) and Vybrant Dil (V22885) were from Thermo Fisher. Drugs for use in infection experiments were dissolved in Hybri-Max dimethyl sulfoxide (DMSO) (D2650; Sigma) to give stock solutions at 1,000-fold their intended final concentration and filter sterilized via 0.2-μm nylon syringe filters. Solvent controls contained 0.125% DMSO.

Compound extraction and LC-MS quantification.

BMDMs were plated on non-tissue culture-treated 6-well plates (0877249; Corning Life Sciences) at 8 × 10⁵ cells/well and grown overnight in BMDM medium at 37°C in humidified 5% CO₂. Cells were mock infected by treatment with infection medium for 4 h or infected as described above. BMDMs were washed twice with warm PBS and then grown in BMDM medium with DMSO vehicle (0.1%) or compounds dissolved in DMSO at 10 μM with or without NH₄Cl (20 mM) to accomplish acidic compartment neutralization. After 24 h, cells were washed twice in warm PBS and detached by incubation with 1 ml/well trypsin-EDTA (0.25%; Gibco) for 10 min at 37°C. Trypsin was inactivated and cells transferred to 15-ml centrifuge tubes by gently aspirating wells twice with 2 ml warm PBS. Cells were collected by centrifugation (514 × g/5 min/25°C) and supernatants were fully decanted.

Analytes were extracted from cell pellets in organic solvent (500 μl acetonitrile-methanol-water [2:2:1]), and 1 μl of freshly prepared 10 mM stocks of clomipramine or clofazimine was used as an internal standard. We disrupted the cells by adding 0.1-mm glass beads and beating in an MP Biomedicals FastPrep-24 homogenizer (two cycles of 30 s at 6.0 m/s). Cellular debris was pelleted by centrifugation (2,800 × g/5 min/25°C) and the organic supernatant collected. A second extraction (500 μl; organic solvent without internal standard) was followed by vigorous vortexing (30 s twice) and centrifugation as described above. The organic extracts were combined and stored at −80°C. Analytes extracted from M. tuberculosis infection experiments were additionally filter sterilized (0.2 μm; polyvinylidene difluoride [PVDF]) twice prior to removal from the biosafety level 3 facility.

Analytes were quantified by UV absorbance (200 nm) on an Agilent 1260 infinity high-performance liquid chromatograph in tandem with an Agilent 6120 quadrupole mass spectrometer. Separation was performed on Agilent Poroshell 120 EC-C₁₈ 2.7-μm columns (4.6 by 50 mm for SRT, kSRT, and FLX; 4.6 by 150 mm for qFLX) at 25°C and a flow rate of 0.4 ml/min. Mobile-phase solvents were H₂O plus 0.1% trifluoroacetic acid (TFA) (solvent A) and acetonitrile plus 0.1% TFA (solvent B).

The eluent profile for SRT and FLX was the following: T₀ = 35% solvent B, T₂₀ = 35% solvent B, T₂₁ = 80% solvent B, T₂₆ = 80% solvent B, T₂₇ = 35% solvent B, and T₃₀ = 35% solvent B. The eluent profile for qFLX, kSRT, and clofazamine had the same time increments but contained 50% solvent B for the isocratic and reequilibration segments. Samples were thawed and passed throughout 0.2-μm filters (PVDF; EMD Millipore), and equal volumes (15 μl) were analyzed for all samples. UV detection was carried out at 200 nm, and m/z was determined in positive mode. Analyte peak retention times and mass signatures were determined from pure standards. Experimental samples were analyzed in duplicate, and concentration was determined from a standard curve run in triplicate in the same analytical sequence as the experimental samples.

Imaging.

BMDM were grown at a density of 1 × 10⁵/well in eight-chambered coverglass culture slides (155411; Nunc) and incubated overnight with 5 μg/ml clofazimine with or without NH₄Cl (20 mM) or bafilomycin A1 (5 nM) as labeled. For clofazimine accumulation imaging, macrophages were washed with warm PBS with calcium and magnesium and incubated at room temperature with Hoechst 33342 at 10 μM for 5 min. Nuclear counterstain solution was removed and replaced with PBS at 4°C, and images were acquired rapidly by wide-field fluorescence microscopy (Axio Observer D1; Zeiss). Images were processed and analyzed by ImageJ software (U.S. National Institutes of Health, Bethesda, MD). For cellular volume studies, cells were grown on glass coverslips, treated as described above, and then stained with Hoechst 33342 (3.25 μM), calcein AM (1 μM), and Vybrant Dil (3.5 μM) for 20 min prior to fixation with 4% paraformaldehyde (15 min). Coverslips were mounted in Vectashield mounting medium (Vector Laboratories) and imaged on a Nikon Ti-E fluorescence microscope coupled with an A1R confocal system (60× objective, 0.5-μm Z-step size). Three-dimensional image reconstruction and cell volume calculations were performed in Imaris (v. 8.2.0; Bitplane) for at least 500 cells from a minimum of eight fields of view distributed across two biological replicates, each containing two technical replicates.

Statistics and chemical property prediction.

All statistical analyses were performed with Prism, version 6.05 (GraphPad Software, San Diego, CA). Marvin was used to draw structures, and calculator plugins were used for structure property prediction and calculation, including CLogP, LogD, and pK_a (Marvin, version 15.5.11.0; ChemAxon; chemaxon.com).