|Home | About | Journals | Submit | Contact Us | Français|
In the last 40 years, only a single new antituberculosis drug was FDA approved. New tools that improve the drug development process will be essential to accelerate the development of next-generation antituberculosis drugs. The drug development process seems to be hampered by the inefficient transition of initially promising hits to candidate compounds that are effective in vivo. In this study, we introduce an inexpensive, rapid, and BSL-2 compatible infection model using macrophage-passaged Mycobacterium tuberculosis (Mtb) that forms densely packed Mtb/macrophage aggregate structures suitable for drug efficacy testing. Susceptibility to antituberculosis drugs determined with this Mtb/macrophage aggregate model differed from commonly used in vitro broth-grown single-cell Mtb cultures. Importantly, altered drug susceptibility correlated well with the reported ability of the respective drugs to generate high tissue and cerebrospinal fluid concentrations relative to their serum concentrations, which seems to be the best predictors of in vivo efficacy. Production of these Mtb/macrophage aggregates could be easily scaled up to support throughput efforts. Overall, its simplicity and scalability should make this Mtb/macrophage aggregate model a valuable addition to the currently available Mtb drug discovery tools.
With one-third of the world population latently infected with Mycobacterium tuberculosis (Mtb), tuberculosis (TB) remains a global health threat. While drug intervention can be successful for the treatment of TB, success is often hampered by long treatment periods lasting over 6 months1 and the requirement to treat with multiple drugs, a scenario that often leads to lack of compliance.
To be effective, drugs must be transported from the blood compartment to nonvascularized pulmonary lesions and diffuse into necrotic foci through multiple cell types.2 At the site of infection, Mtb remains hidden inside macrophages and/or encapsulated within granuloma structures, a hallmark feature of Mtb infection.3 Finally, drugs have to cross the highly impermeable, lipid-rich cell envelope of Mtb,4 and need to reach their molecular target at adequate concentrations. To further complicate matters, Mtb infection often persists long term in a hyperdrug-resistant, latent state.5 This poses a problem to standard anti-TB drugs since the vast majority of these only target metabolically active Mtb. Thus, both tissue penetration capacity and efficacy against latent Mtb are two critical parameters that define TB drug potency. A major problem with the successful transfer from discovery to application for Mtb drugs is that these parameters have not been given sufficient consideration during in vitro Mtb drug screening or the early lead development process.6,7 The majority of drug screens were performed using extracellular in vitro broth-grown single-cell culture assays, in which Mtb is actively growing and only the Mtb cell membrane acts as a potential diffusion barrier.
More recently, this problem has been recognized and drug screening assays against nonreplicating, latent Mtb8,9 or intracellular Mtb have been established,10–12 but even these more advanced assays do not consider the penetration barriers that drugs encounter in the nonvascularized pulmonary lesions, or the necrotic foci at the site of infection. The ability to examine direct tissue penetration of drugs is only achieved using advanced experimental approaches and high-end instrumentation such as matrix-assisted laser desorption/ionization mass spectrometry imaging.13–15 These highly advanced methods have been used to generate drug penetration profiles in in vivo lung lesions and granulomas,13 but are limited to a few laboratories in the world due to specific instrumentation requirements, and as such are prohibitively expensive to be applied during the early drug development process.
Even a low-throughput drug susceptibility assay that could simultaneously indicate the efficacy of candidate compounds to diffuse through physiologically relevant penetration barriers and exert activity against latent Mtb would significantly add to the currently available research tools to prioritize discovered hit compounds.
THP-1 monocytes (ATCC TIB-202) were maintained in the RPMI 1640 medium supplemented with 2mM l-glutamine and 10% heat-inactivated fetal bovine serum (FBS) at 37°C in a humidified atmosphere of 5% CO2. FBS was obtained from Life Technologies. Resazurin, rifampicin, streptomycin, moxifloxacin, ciprofloxacin, gentamicin, clarithromycin, and cycloserine were purchased from Sigma-Aldrich. Hygromycin B was purchased from Calbiochem.
The M. tuberculosis H37Rv-derived auxotroph strain mc26206 was grown in the Middlebrook 7H9 medium (Difco) supplemented with 0.2% glycerol, 0.02% tyloxapol, and 10% OADC (Remel) or on Middlebrook 7H10 plates supplemented with 0.5% glycerol and 10% OADC. Growth media of the auxotrophic M. tuberculosis strain were supplemented with 24μg/mL pantothenate and 50μg/mL l-leucine.16 The plasmid pMN43717 was transformed into M. tuberculosis mc26206 to express GFP, which was used for all experiments in this study. Nonreplicating nutrient-starved Mtb was cultured by replacing the normal Mtb media of a culture at OD600=1.0 with phosphate-buffered saline, 0.02% tyloxapol, 24μg/mL pantothenate, and 50μg/mL l-leucine. Cultures were left standing at 37°C for 4 weeks before drug susceptibility testing. Single-cell Mtb for drug susceptibility testing was obtained by water bath sonication (Fisher FS-60; 130W; 3×10s pulses) of broth-grown Mtb cultures.
M. tuberculosis mc26206 growing in log-phase was quantified by optical density measurement at 600nm using the conversion of 3×108 bacteria per mL for OD 1.0. The amount of bacteria required to achieve a multiplicity of infection (MOI) of 25 were washed, resuspended in cell culture media, and added to the THP-1 cultures. Monocytes and Mtb were then resuspended together to ensure even exposure of the monocytes to Mtb. Media were changed every 2–3 days during an infection time of 10–14 days.
Mtb-infected THP-1 aggregates from 10- to 14-day old cultures were harvested and resuspended in Mtb growth media. As the strain of Mtb used in this model neither replicates nor is killed during this infection period, we adjusted the bacteria concentration to 6 million/mL based on initial Mtb input on the day of infection and the total culture volume during harvest. At the same time, in vitro broth-grown single-cell Mtb cultures were adjusted to a final bacteria concentration of 6 million/mL corresponding to an optical density at 600nm of 0.02 (using the conversion of 3×108 bacteria per ml for OD 1.0). Mtb derived from Mtb/macrophage aggregates (Mtb-aggregate) and in vitro broth-grown single-cell cultures (Mtb-single) were distributed to 96-well plates in a final volume of 200μL containing antibiotics in a range of concentrations. The 96-well plates were incubated for 5 days at 37°C. Then, 20μL of resazurin dye mixture was added to achieve a final concentration of 100μM resazurin and 0.5% Tween 80 and incubation at 37°C was continued for another 16–30h before the plates were analyzed by a Synergy HT plate reader (BioTek). Conversion of the indicator dye resazurin into resorufin with a fluorescence emission peak at 590nm served as a marker for bacterial growth. Refer to Table 1 for detailed protocol.
M. tuberculosis bacilli were extracted from Mtb/macrophage aggregates by water bath sonication for 5min. Then, Mtb was smeared onto glass slides and heat fixed. Nile Red (Acros Organics) was dissolved in ethanol and used at a working concentration of 10μg/mL to cover the fixed Mtb on slides for 15min at room temperature. Slides were gently washed with water and mounted with a coverslip in FluorSave mounting solution (Millipore).
For Nile Red staining, imaging was performed using an Axiovert 200 microscope (Carl Zeiss) equipped with a 100×/1.4 Plan-Apochromat (Carl Zeiss) and the filter set 49004-ET-CY3/TRITC (Chroma). Images were recorded using an AxioCam MRc camera (Zeiss) coupled to Axiovision v4.5 software (Carl Zeiss). For automated imaging of Mtb/macrophage aggregation within 96-well plates, the Cytation 3 Cell Imaging Multi-Mode Reader (BioTek) fitted with 2.5×or 20×objectives and the GFP filter set were used. To examine Mtb/macrophage aggregates, PE-conjugated antibody to ICAM-1 (BD Biosciences) and the cell viability staining dye 7-amino-actinomycin D (7AAD; eBioscience) were used at a dilution of 1:20 according to manufacturer's protocol. Thereafter, cells were imaged with the EVOS FL Cell Imaging System (Life Technologies) with the combination of GFP/RFP light cubes at 20×magnification.
To provide an efficient experimental tool for improved in vitro Mtb drug development, we established a simple Mtb infection model that is amenable for early hit verification or even drug screening. For reasons of cost efficiency and general usability, it is highly desirable to use a mycobacteria strain that is suitable for work under BSL-2 conditions, but this strain needs to remain as closely related to the model strain, Mtb H37Rv, as possible. While the commonly used Mtb surrogates M. bovis BCG and M. smegmatis satisfy the BSL-2 condition, these strains are not closely related to Mtb. BCG is an attenuated vaccine strain derived from M. bovis that does not naturally infect humans, while M. smegmatis is a rapid-growing mycobacterium that lacks most Mtb virulence genes and has different antibiotic resistance profiles compared to Mtb. Thus, we chose to use Mtb mc26206 (ΔpanCD, ΔleuCD), a specific derivative of H37Rv16 that retains all but four genes, panCD and leuCD, which are neither implemented in virulence nor drug resistance. The auxotrophic Mtb mc26206 strain is a BSL-2 organism that can be handled in almost any laboratory. Its utilization massively reduces experimental time investment and costs compared to studies utilizing Mtb H37Rv, which requires handling in BSL-3 facilities. Mtb mc26206 maintains the RD-1 locus (absent from BCG), which plays a major role in Mtb virulence and granuloma formation.18 Importantly, we determined the minimal inhibitory concentrations (MIC) of standard TB drugs on Mtb mc26206 and show that the strain's requirement for supplementation with l-leucine and pantothenate to replicate does not affect its drug susceptibility profile compared to the parent H37Rv strain (Table 2). Furthermore, auxotrophic Mtb strains have already been used in well-established drug evaluation studies for Mtb.19–21 For our experiments, we utilized a gfp-expressing strain of Mtb mc26206 to infect THP-1 cells, a monocyte cell line widely used in Mtb research.22,23 The ability of this Mtb strain to express GFP allows for the visual control of macrophage phagocytosis efficacy and direct quantification of infection levels. Since this auxotrophic strain does not replicate during macrophage infection, we rationalized to use a higher than standard MOI to mimic the intracellular replication observed with wild-type Mtb, where the bacterial burden is known to increase by 10-fold after ~5 days of infection.24
Infection of nondifferentiated THP-1 monocytes at an MOI of 25 initially showed features of limited differentiation to a macrophage-like phenotype as indicated by macrophage aggregation after 3–4 days (Fig. 1A). THP-1 cells did not differentiate into the classic phorbol 12-myristate 13-acetate (PMA)-induced macrophage phenotype and adhered to the bottom of the wells, but rather started to increase in size and adhered to neighboring cells, thereby forming Mtb/macrophage aggregate structures. By day 8–10, the size of Mtb/macrophage aggregates built around a densely packed core of Mtb had increased to >100μm (Fig. 1B). Mtb/macrophage aggregates could be directly generated in 96-well plates (Fig. 1C–E) or the production could be efficiently scaled up to larger flask volumes (e.g., T-75 flasks) to produce sufficient material for throughput-format experiments.
The underlying idea of creating large Mtb/macrophage aggregate structures to study drug efficacy was to generate a biologically relevant model that resembles in vivo conditions, where Mtb is trapped and embedded within multiple layers of immune cells and necrotic tissue, which alter the penetration efficacy of drugs. To examine the biological relevance of this model, we probed the experimental system for the presence of a (i) central necrotic core, (ii) the containment of Mtb within the core of the aggregates by viable macrophages, (iii) a possibly latent Mtb phenotype, and (iv) the presence of viable Mtb residing inside macrophages. All of these parameters are important features at sites of Mtb infection in vivo.5,25,26
The core of the Mtb/macrophage aggregates contained densely packed Mtb (Fig. 1B) and dead cell material that stained positive for 7AAD (Fig. 2A), reproducing previously reported observations that Mtb-infected human macrophages form cell aggregates that show prominent features of necrosis, such as the release of DNA into the extracellular space.27 The aggregate core was enclosed by a layer of noninfected macrophages that expressed high levels of ICAM-1 (Fig. 2B), an adhesion molecule known to play an essential role in in vivo granuloma formation,26 where increased ICAM-1 expression during Mtb infection is necessary to promote cellular adhesion.28 Functionality of the enhanced expression of surface adhesion molecules indicates tight cell–cell interactions, as confirmed by the fact that even mechanical stress (e.g., rigid pipetting) did not result in relevant disintegration of the Mtb/macrophage aggregates. Agitation released single viable macrophages that harbored intracellular Mtb, which could be isolated from the aggregate structures (Fig. 2C). Outgrowth experiments from aggregate-derived Mtb revealed that almost 100% of the Mtb cells harvested from 11-day-old cultures were viable when compared to the inoculum on day 1 (Fig. 2D). Furthermore, Mtb extracted from the highly dense Mtb/macrophage aggregate core stained positive with the lipophilic dye Nile Red (Fig. 2E), which detects the presence of intracellular lipid bodies suggesting the possibility of latent mycobacteria.29–31 At the least, this result indicated metabolic changes of Mtb following passage through these aggregate forming macrophage infection cultures.
Drug penetration into lung tissues, the primary site of Mtb infection and granuloma formation, is complex and involves different processes such as passive diffusion, permeation, active transport, and bulk flow.32 Accordingly, various classes of antibiotics show different penetration properties. For instance, aminoglycosides are too polar to diffuse directly across lipid membranes, and therefore enter cells slowly by endocytosis.33 Uptake of β-lactam antibiotics is also poor in alveolar macrophages.34 Conversely, uptake of rifampicin, tetracycline, or ethambutol into alveolar macrophages is relatively faster, and quinolone levels in lung tissue are often 15–20 times higher than serum levels.35 These pharmacological characteristics define in vivo drug efficacy, but are inadequately addressed by most other standard evaluation methods.
To explore whether the Mtb/macrophage aggregate model could improve our ability to predict the potential in vivo efficacy of hit or lead compounds, we tested the efficacy of several standard Mtb drugs against Mtb in Mtb/macrophage aggregates relative to their efficacy on in vitro broth-grown single-cell Mtb cultures. For this purpose, we used a panel of commonly used antibiotics and TB drugs for which properties such as penetration capacity into lung tissue or cerebrospinal fluid (CSF),36–38 as well as activity against persistent Mtb39 were reported (rifampicin, moxifloxacin, ciprofloxacin, streptomycin, gentamicin, cycloserine, and clarithromycin). Mtb/macrophage aggregates were harvested from large-scale infection cultures after 14 days, while in vitro broth-grown single-cell Mtb cultures were used as references. Thereafter, Mtb/macrophage aggregates or in vitro broth-grown single-cell Mtb were incubated in mycobacteria growth media (7H9 media) containing a titration of the aforementioned drugs for 5 days. As the purpose of this study is to test drug efficacy on macrophage-passaged Mtb aggregates that should more closely resemble in vivo conditions, the use of mycobacteria growth media is essential for compatibility with a growth inhibition assay to measure drug effects. Thus, we chose to use the well-characterized resazurin microtiter assay for mycobacteria growth40 to quantify the drug susceptibility of Mtb derived from our Mtb/macrophage aggregate model in 96-well microplate format. The eventual disintegration of macrophages in mycobacteria growth media releases Mtb aggregates into the medium and allows Mtb cells to undergo enough bacterial replication to enable sensitivity for measurement by the resazurin assay. The addition of 0.5% Tween 80 at the beginning of the resazurin incubation period would eliminate all remaining macrophages in the sample, and thus ensure that the measured metabolic activity would only be associated with Mtb that survived the antibiotic treatment. This was confirmed by the fact that noninfected THP-1 cells treated in the same manner did not register any metabolic activity (data not shown).
Suboptimal dosing of rifampicin is a widely recognized problem.41–43 Rifampicin, despite being the antibiotic of choice to treat Mtb infection, is known to ineffectively penetrate pulmonary granulomatous lesions,14 or only reach low concentrations in the CSF of tuberculous meningitis patients.38 Consistent with these reports, the MIC of rifampicin was >10-fold higher in Mtb from Mtb/macrophage aggregate cultures compared to in vitro broth-grown Mtb control cultures, as determined by susceptibility killing curves (Fig. 3A and Table 3). A possible explanation for this finding could be that the large size (>800Da) and lipophilic properties37 may result in incomplete rifampicin penetration through the many layers of cell membranes and debris that create the Mtb/macrophage aggregate structures, and particularly into the necrotic core of the aggregates.
To test the idea that the Mtb/macrophage aggregates could affect drug penetration, we next tested the ability of the fluoroquinolone moxifloxacin to exert its anti-Mtb activity in this model. Moxifloxacin was reported to penetrate granulomatous lesions more effectively than rifampicin.14,15 Similar to rifampicin, moxifloxacin provides satisfactory activity against persistent Mtb in in vitro models.39 Differences in the relative ability of these two antibiotics to act against Mtb derived from macrophage aggregates or in vitro broth-grown cultures should thus be mostly attributable to tissue diffusion characteristics. Mtb in the Mtb/macrophage aggregate cultures indeed exhibited a 4- to 8-fold relative increase in tolerance to moxifloxacin (Fig. 3B and Table 3). Similar findings were made for ciprofloxacin, another fluoroquinolone (Fig. 3C and Table 3). Increased tissue penetration compared to rifampicin could possibly be the result of different pharmacological properties, with fluoroquinolone compounds exhibiting moderate lipophilicity and being of lower molecular weight (~300Da).37
Streptomycin, a first-line TB drug belonging to the class of aminoglycoside antibiotics that are known to poorly diffuse across membranes,33 is relative large (>500Da) and hydrophilic,37 properties that hinder penetration into the CSF.36,38 These properties correlated well with the fourfold increase in MIC to streptomycin for Mtb from aggregate cultures, relative to in vitro broth-grown single-cell cultures (Fig. 3D and Table 3). Although aminoglycosides generally exhibit poor tissue penetration, gentamicin concentrations in the alveolar lining fluid of patients were found to correlate well with serum concentrations with a penetration ratio of 0.32.44 Consistent with this physiological finding and not with its chemical properties, Mtb in Mtb/macrophage aggregate cultures showed no difference in susceptibility to gentamicin, relative to in vitro broth-grown Mtb (Fig. 3E and Table 3).
Last, consistent with its reported excellent CSF penetration properties36 and small size (~100Da), cycloserine sensitivity was unchanged against Mtb in Mtb/macrophage aggregate cultures when compared to in vitro broth-grown Mtb (Fig. 3F and Table 3). Also, susceptibility of Mtb from aggregate cultures to clarithromycin, a macrolide antibiotic, was similar or even slightly increased relative to in vitro broth-grown Mtb cultures (Fig. 3G and Table 3), a result that is consistent with its pharmacokinetic properties that show high tissue concentrations relative to serum concentrations, indicating effective diffusion into lung tissue and phagocytes.32,45 In summary, these results suggest that the data obtained using the Mtb/macrophage aggregate model are likely good predictors of drug efficacy in a physiologically relevant setting where tissue penetration is critical.
As revealed by Nile Red stains and CFU enumeration experiments (Fig. 2D, E), the introduced Mtb/macrophage aggregate model produces nonreplicating Mtb that has accumulated large amounts of lipid bodies, features suggestive of a latent infection phenotype.31 To investigate whether it is possible to determine the relative contributions of (i) a latent Mtb infection status29,46 or (ii) the diffusion barrier of the core structure, to the increased drug tolerance of Mtb observed in the Mtb/macrophage aggregate model, we determined drug susceptibility killing curves against nonreplicating latent Mtb cells generated by the commonly used nutrient starvation method.9,47 In these latent Mtb cultures, tolerance to the four drugs (rifampicin, moxifloxacin, ciprofloxacin, streptomycin) that showed increased MICs against Mtb from Mtb/macrophage aggregates were relatively unchanged with the exception of rifampicin, which showed a modest two to fourfold increase compared to broth-grown Mtb cultures (Fig. 4A–D and Table 3). This result is consistent with reports of increased rifampicin resistance in long-term persistent Mtb that show features of latency or nonreplicating phenotype.9,29,46 Furthermore, this suggests that the >10-fold increase in the MIC of rifampicin against Mtb from Mtb/macrophage aggregates is mainly modulated by the altered drug diffusion properties of the Mtb/macrophage aggregates, and only to a lesser extent by a latent infection status.
As with most throughput assays, the Mtb/macrophage aggregate model is a simplification and reductionist approach to an otherwise complex biological system, and thus would not suffice as a complete granuloma model. However, the following key features of the Mtb/macrophage aggregate structures: (i) a highly dense necrotic core containing Mtb with a high content of intracellular lipid bodies29–31 (Fig. 2E), (ii) the presence of viable macrophages that harbor intracellular, viable Mtb (Fig. 2C, D), and (iii) the tight, multilayer envelope of noninfected macrophages surrounding the infected core of the aggregate structure (Fig. 1) were sufficient to produce in vitro drug efficacy data with very good correlation to reported in vivo tissue penetration data.14,15,38,42 Drugs like cycloserine, clarithromycin or gentamicin, known to effectively penetrate into tissues, exhibited no efficacy differences in the Mtb/macrophage aggregate model when compared to in vitro broth-grown Mtb cultures, while the reduced tissue penetration capacity of rifampicin or moxifloxacin was efficiently predicted by the Mtb/macrophage aggregate model (Fig. 3A, B).
Drug susceptibility of intracellular Mtb differs from in vitro broth-grown Mtb as it has been shown that increased drug concentrations are necessary to be effective against intracellular mycobacteria.43,48 Our data are in strong agreement with these findings showing that rifampicin, streptomycin, and fluoroquinolones require higher concentrations to inhibit growth of intracellular Mtb.48 Our model further demonstrates that rifampicin has the largest discrepancy in efficacy against Mtb from Mtb/macrophage aggregates compared to in vitro broth-grown Mtb, while streptomycin and fluoroquinolones show smaller differences, all of which are consistent with previously established drug efficacy testing against intracellular Mtb.48 This would suggest that our model is indeed reflective of or even comparable to other intracellular Mtb drug susceptibility infection models. In addition to these valuable and informative high-content screening methods against intracellular Mtb,10–12,48 recent development of an in vitro granuloma-based Mtb drug susceptibility phenotyping confirms that drug concentration requirements are altered within granulomas,49 which further supports findings from our simplified Mtb/macrophage aggregate model. Therefore, that our model is able to produce drug efficacy data consistent with several intracellular Mtb infection models shows its promise as an alternative method to quickly and reliably assess drug efficacy against Mtb during the early drug development process. Importantly, our infection model is advantageous in (i) its ability to generate unlimited amounts of Mtb/macrophages aggregates (in vitro granulomas) necessary for high-throughput screening compared to using primary macrophages from donors, (ii) is BSL-2 compatible, thus saving valuable time and money, and (iii) does not require high-end instrumentation compatible with high-content imaging analysis.
As such, our alternative model would address one of the key problems in Mtb drug development, which is the inefficient correlation of in vitro drug efficacy of currently used screening methods7 with in vivo outcomes. Our data suggest that this disconnect would be, in particular, a problem for the ability of some screening models to predict drug penetration efficacy of the candidate compounds into the granulomatous or necrotic tissues at the site of Mtb infection. The generation of truly authoritative and accurate drug penetration profiles in in vivo lung lesions and granulomas13–15 will remain limited to a few laboratories in the world, and due to specific instrumentation requirements, these methods are not amenable to throughput analysis early in the drug development process. Our model would provide a cost-effective, rapid, and BSL2-compatible alternative with good predictive capability for the in vivo ability of antibiotics to penetrate into key tissues such as the lung or the CSF.32,36–38,45
We thank Dr. William Jacobs (Albert Einstein University, New York) for kindly providing M. tuberculosis mc26206. This work was funded, in part, by NIH grants R01-AI104499 and R21-AI116188 to O.K. and R01-AI104952 to F.W. Parts of the work were performed in the UAB CFAR BSL-3 facilities and by the UAB CFAR Flow Cytometry Core/Joint UAB Flow Cytometry Core, which are funded by NIH/NIAID P30 AI027767 and by NIH 5P30 AR048311.
A.S., F.W., M.N., O.K., and J.S. conceived and designed experiments and analyzed data; K.S., V.H., A.S., and J.S. performed experiments; and F.W., M.N., O.K., and J.S. wrote and edited the article.
No competing financial interests exist.