Confinement-Induced Drug-Tolerance in Mycobacteria Mediated by an Efflux Mechanism

Tuberculosis (TB) is the world’s deadliest curable disease, responsible for an estimated 1.5 million deaths annually. A considerable challenge in controlling this disease is the prolonged multidrug chemotherapy (6 to 9 months) required to overcome drug-tolerant mycobacteria that persist in human tissues, although the same drugs can sterilize genetically identical mycobacteria growing in axenic culture within days. An essential component of TB infection involves intracellular Mycobacterium tuberculosis bacteria that multiply within macrophages and are significantly more tolerant to antibiotics compared to extracellular mycobacteria. To investigate this aspect of human TB, we created a physical cell culture system that mimics confinement of replicating mycobacteria, such as in a macrophage during infection. Using this system, we uncovered an epigenetic drug-tolerance phenotype that appears when mycobacteria are cultured in space-confined bioreactors and disappears in larger volume growth contexts. Efflux mechanisms that are induced in space-confined growth environments contribute to this drug-tolerance phenotype. Therefore, macrophage-induced drug tolerance by mycobacteria may be an effect of confined growth among other macrophage-specific mechanisms.


Introduction
Tuberculosis, caused by infection with Mycobacterium tuberculosis (Mtb) remains one of the world's deadliest diseases, killing an estimated 1.5 million people annually [1]. Whereas drugsusceptible forms of the disease are in principle curable, the duration of treatment courses is at least 6 months and may last years [2]. Multidrug resistant TB (MDR-TB) and extensively drug resistant TB (XDR-TB) have poorer and less certain outcomes [2][3][4]. It is expected that shortened antituberculosis treatment regimens will improve patient adherence to treatment, and thereby foster better case management and disease control and minimize the risk of drug resistance [5,6]. An interesting aspect of long-term chemotherapy in TB is that, whereas more than 95% of the tubercle bacilli population detectable in a patient's sputum can be cleared in the first few days of treatment, prolonged treatment is required to eradicate the residual minority population (<5%) [7,8] using drugs that are rapidly potent in vitro [9][10][11]. Overcoming this bacterial persistence is central to shortening TB treatment [12]. The transient nature of persistence-an epigenetic drug-tolerance phenotype-has pushed research in this area toward miniaturization and chip-based control using time-lapse microscopy to study the phenotypic heterogeneity within bacterial populations in situ [13][14][15]. Whereas these efforts have thus far investigated persistence in an extracellular context, recent studies have shown that the intracellular (or intramacrophage) mycobacterial sub-population, which makes up an essential component of human TB infection [16,17], is significantly more tolerant to antibiotics [18,19]. Separate studies have shown that the dimensions and diffusional characteristics of the growth environment can influence bacterial gene expression [20][21][22]. To investigate persistence, we characterized the growth and drug susceptibility of mycobacteria replicating in space-confined microfabricated cell culture environments (or microdialysers) to mimic the confinement experienced by mycobacteria replicating within macrophages.
We focused our experiments on Mycobacterium smegmatis, an experimentally tractable surrogate for M. tuberculosis with respect to rifampicin-a frontline drug in TB treatment [23]. The microdialyser has a micro-sized cell culture chamber that is 200 picoliters (pL) in volume to approximate the~5pL volume of the membrane-bound compartment of human macrophages [24]. Thus upon inoculation of a microdialyser culture chamber with~5 mycobacteria, the cell density immediately exceeds 10 7 cells/ml and verges toward the effective cell density of an intra-macrophage mycobacterium,~10 8 cells/ml [24].

Microfluidic device design and fabrication
The microdialyser chip was fabricated out of the silicone elastomer polydimethylsiloxane (PDMS) (General Electric RTV 615) using multi-layer soft lithography, as described previously [25]. Up to 120 microdialyser units can run in parallel on each chip.

The microdialyser reader
Mycobacteria were cultivated in growth chambers within the microdialyser chip (S1 Fig) which was positioned for live-cell imaging on an Olympus IX81 inverted microscope furnished with a PRIOR Scientific XYZ motorized stage system (Wirsam Scientific Precision Equipment (Pty) Ltd., Pinetown, South Africa). The motorized stage system enabled documentation of multiple simultaneous microdialyser cultures on a single chip. Imaging was done using a Plan Fluor 40X 0.6NA objective. Digital images were captured by a Hamamatsu digital CCD ORCA-R2 camera (Wirsam Scientific Precision Equipment (Pty) Ltd., Pinetown, South Africa). LabVIEW software was used to control the synchronized operation of these components and chip valve actuation.

Media, strains and growth conditions
M. smegmatis strain mc 2 155 was received from Bill Jacobs (Albert Einstein College of Medicine). M. smegmatis ΔftsEX-gift from Eric Rubin (Harvard University) comprised of an M. smegmatis mc 2 155 strain with the ftsEX gene deleted and were hypersensitive to the RNA polymerase inhibitor, rifampicin, with a minimum inhibitory concentration of 1μg/ml [26]. E. coli Top10F' cells were purchased from Invitrogen. Conventional cell cultures were performed with 10 ml of 7H9 broth in tissue culture flasks at 37°C with shaking at 100 rpm and growth was determined by measuring the turbidity of the cultures at 600 nm (OD 600 ) twice daily unless otherwise stated. M. smegmatis cells in the microdialyser were grown in 7H9 broth at 37°C. The M. smegmatis mmpL11 mutant was described previously [27].
M. smegmatis precultures were prepared by inoculating a 10ml medium (7H9 broth) sample with cells from a frozen stock and culturing at 37°C with shaking at 100 rpm until an OD 600 reading of 0.8 was reached. The cell culture was centrifuged at 2500 rpm for one minute to cause the large cell clusters to gather towards the bottom of the tube. To load the cells into the microdialyser chip, first~500μL of the supernatant cell suspension was aspirated into a piece of tygon tubing (0.02 OD X 0.06 ID, Cole Parmer) by suction using a 1 ml syringe. Next, the free end of the tygon tubing was connected to a cell input port on the microdialyser chip (see S1 Fig), using a stainless steel pin as an adaptor between the chip and the tygon tubing. The cells were introduced into each growth chamber by flowing the cell suspension from the tygon tubing through each chamber: in via an open inlet and out via an open outlet. Accordingly, cells were trapped in each growth chamber once the inlet and outlet of the chamber was subsequently closed, by actuating the appropriate valves on the chip. On average, cells were trapped in each chamber at a uniform cell density of~2×10 7 cells/ml. Thus, on average, the starting number of cells per growth chamber in the 200, 500, 1200 and 1700pL growth chambers was 5, 12, 30 and 42 cells respectively. Upon microdialyser inoculation, cells were incubated in the microdialyser growth chambers by maintaining the chip at 37°C using a Solent Scientific microscope incubation chamber heating system (Wirsam Scientific Precision Equipment (Pty) Ltd., Pinetown, South Africa).

The microdialyser operation process
The microdialyser uses a microdialysis scheme to periodically introduce fresh nutrients and remove waste from a captive population of mycobacteria through diffusive exchange, mimicking the passive bidirectional exchange of pro-and anti-mycobacterial factors across the macrophage membrane (Fig 1). The 200pL mycobacterial growth chamber of the microdialyser is connected via one or more link valves to an adjacent conditioning chamber that stores fresh medium ( Fig 1A). Periodically (typically, hourly), the link-valves are opened for 60 seconds to allow small molecules to freely diffuse between the two chambers down their respective concentration gradients: fresh nutrients diffuse into the growth chamber while mycobacterial metabolic waste products diffuse into the conditioning chamber ( Fig 1B). The relatively large size and non-motility deter the mycobacterial cells from exiting the growth chamber when the link valves are opened. Next, the conditioning chamber is refilled with fresh medium and awaits the next microdialysis step. We calibrated the diffusive exchange functionality of the microdialyser using colorimetric assays and demonstrated that on average, the microdialyser replaced the growth chamber fluid within six microdialysis steps ( Fig 1C).

Colorimetric assays
The food dye used in the colorimetric microdialyser characterization assay was obtained from McCormick & Co., Hunt Valley, MD. The dye concentration in the microdialyser was determined based on the average pixel value of optical micrographs of a region within the growth chamber or the conditioning chamber during a series of microdialysis steps. The camera pixel value (P) is linear with respect to transmission (T), which is the anti-log of the negative of the optical density (OD) [28], as depicted in the equations below.

Microscopic cell surface density
The microdialyser architecture is such that all the cells dwell in a chamber 10 μm high, equivalent to the focal depth of the Plan Fluor 40X 0.6NA objective. We assessed bacterial growth in the microdialyser by enumerating the fraction of pixels occupied by M. smegmatis cells in images of microdialyser culture chambers (or cell surface density-a dimensionless quantity with a maximum value of 1), at different time points. We developed image-processing algorithms written in Matlab to determine the cell surface density in each picture. The motorized stage system enabled documentation of multiple simultaneous microdialyser experiments on a single chip.

Results and Discussion
At full operational capacity, the microdialyser chip can support 120 microdialyser culture units that operate independently, with each culture monitored in situ by optical microscopy to provide automated, real-time, non-invasive measurement of cell density. Mycobacterial growth was assessed from microscopic images of each growth chamber by enumerating the fraction of pixels occupied by bacterial cells (or cell surface density-a dimensionless quantity with a maximum possible value of 1) at different time points. Microdialyser growth curves followed the trend of typical mycobacterial growth: upon inoculation, a typical culture began with a lag period, followed by an exponential growth phase that gave way to a stationary phase. To functionally validate the effectiveness of the microdialyser in modulating the growth environment of its captive mycobacterial population, we demonstrated the ability to speed up or slow down the mycobacterial growth by switching the growth chamber medium from nutrient-poor to nutrient-rich and vice versa (Fig 2).
Conventional liquid phase cultures of M. smegmatis mc 2 155 cells were treated with a series of antimycobacterial drugs, including rifampicin (35μg/ml), isoniazid (40μg/ml), ofloxacin (100 μg/ml) and hygromycin (100μg/ml). Growth was inhibited in conventional cultures by these drugs at the indicated concentrations ( Fig 3A). Growth in microdialyser cultures was inhibited by the drugs with the exception of rifampicin, which had minimal growth inhibitory effect even at 350 μg/ml-which is 10× the concentration that inhibited growth in the conventional liquid phase cultures,~40× the minimum inhibitory concentration (MIC) and~10× the minimum bactericidal concentration (MBC) of rifampicin for mc 2 155 cells [29,30] (Fig 3B). For comparison, rifampicin at 350 μg/ml inhibited growth of Escherichia coli cells in microdialyser cultures and maintained its antibiotic potency for more than 200 hours, demonstrating that there was sufficient penetrance of the rifampicin into the microdialyser reactors (S2 Fig). This result also indicated that the rifampicin tolerance phenotype was specific to M. smegmatis and absent in E. coli. Given the small number of M. smegmatis cells present in the growth chamber (not exceeding 20) at the time the rifampicin resistance first appeared, a simple mutation rate versus population size argument excludes the possibility of a mutational cause of resistance to rifampicin. Because rifampicin is a front line drug in TB treatment, we sought to better characterize and further elucidate mechanisms underlying the rifampicin tolerance phenotype.
To investigate the role of confinement in the rifampicin resistance of microdialyser cell populations, we fabricated a new chip with growth chambers of various sizes: 200pL, 500pL, 1200pL and 1700pL (S1 Fig). Colorimetric assays ascertained that the diffusive penetrance of the microdialyser process was similar across all growth chamber sizes (S5 Fig). In addition, we obtained an ΔftsEX mutant of M. smegmatis that is particularly hypersensitive to rifampicin with a MIC of 1μg/ml [26]. M. smegmatis ΔftsEX cells grew similarly well in the various growth chamber sizes in drug-free medium. However, although M. smegmatis ΔftsEX demonstrated resistance to rifampicin at 350 μg/ml in the 200pL cultures, the drug inhibited growth in the larger (500pL, 1200pL and 1700pL) cultures (Fig 4). Thus, the rifampicin resistance phenotype was dependent on the size of the growth chamber: appearing when the mycobacteria were cultured in the smallest (200pL) growth chambers and disappearing in the bigger reactor volumes. Wild type mc 2 155 M. smegmatis cells growing in various sized growth chambers had a similar pattern of drug tolerance behavior when exposed to 350μg/ml of rifampicin (S4 Fig).
One potential cause of volume-dependent rifampicin tolerance may be due to increased bacterial density in space-confined environments through quorum sensing. In the generic model of quorum sensing [31,32], bacterial cells secrete signaling molecules called autoinducers whose concentration in the surrounding medium increases with cell density. At low cell density, the autoinducer molecules produced at a basal level diffuse out of the cell, and are ultimately lost to the environment. As the cell density increases, the autoinducer concentration reaches a threshold concentration and triggers a transcriptional response that results in increased expression of virulence determinants [33,34], upregulation of biofilm formation [35][36][37] or entry into stationary phase [38,39]-that is unattainable with low cell density. Large volume cultures require a proportionately large cell population to elicit quorum sensing , the rich medium in cultures 3 and 4 was replaced with nutrient-poor medium using microdialysis, which slowed down growth in these cultures relative cultures 1 and 2 that remained rich. Conversely, at point A, the nutrient-poor medium in cultures 5 and 6 was replaced with rich medium, resulting in faster growth in these cultures relative cultures 7 and 8 that remained nutrient-poor. Cultures 1 to 8 were cultivated simultaneously on the same chip at 37°C. Each condition was demonstrated in at least three cultures. Bottom panels (a to d) show typical area maps of the cells in the growth chamber of microdialyser 1 at the corresponding points from which the cell surface density is determined. Scale bar, 50μm. behavior. However, in a sufficiently small growth environment, a few cells can elicit quorum sensing behavior more efficiently because the autoinducer molecules produced are kept in   close proximity by the boundaries of the confined growth compartment and can therefore accumulate faster [40][41][42]. In a pattern consistent with quorum sensing control, when the time interval between consecutive microdialysis steps was doubled to two hours, the larger microdialyser cultures became more tolerant to rifampicin in a volume-dependent fashion, whereby the fraction of cultures with positive growth increased as the culture volume decreased (Fig 5). Notably, an inevitable consequence of the slower dilution rate was a reduced cellular growth rate due to the proportionately less frequent infusion of fresh nutrients. Bacteria growing with slower growth rates may be more tolerant to rifampicin, and therefore these cells experiencing a slower dilution rate may have a higher frequency of recovery. We measured the bacterial growth rates in the various growth chamber sizes in drug-free medium and found them to be similar (S6 Fig). Therefore, whereas slower growth rate may be a probable cause of the higher recovery frequency at the slower dilution rate, it does not explain the differential drug tolerance in the various growth chamber sizes. This pattern of expression is consistent with quorum sensing and may allow us to further study this system in mycobacteria.
Studies show that microbial efflux pumps can confer epigenetic resistance to antibiotics by enabling bacterial cells to extrude the drug molecules intended to kill them [18,[43][44][45][46]. We explored efflux activity as a potential underlying mechanism for rifampicin tolerance in the 200pL using the efflux inhibitor approach, which is widely used to indicate efflux activity [18,[47][48][49]. M. smegmatis ΔftsEX cells were cultured with rifampicin in the presence of verapamil -a mycobacterial efflux inhibitor [50]. As per the efflux inhibitor approach, dramatically reduced growth of the cells in the presence of verapamil suggested that confinement-induced drug tolerance was mediated by efflux mechanisms (Fig 6).
The discrepancy in drug sensitivity between the small and large volume microdialyser cultures can be exploited to investigate the genetic markers underlying confinement-induced drug tolerance. As a proof of principle, we investigated MmpL11, a cell wall lipid transport protein that contributes to biofilm formation in M. smegmatis [51]. In drug-free 200pL cultures, an M. smegmatis mutant lacking mmpL11 functionality grew in all 12 reactors (Fig 7). However, only 11 of 16 mutant cultures registered growth in rifampicin medium, notwithstanding that the growth rate was slower in these cultures. Loss of MmpL11 function reduced resistance to rifampicin in the 200pL cultures, suggesting that MmpL11 protein contributes to confinementinduced rifampicin tolerance. It should be noted that there is no difference in rifampicin MIC between the wild-type M. smegmatis and mmpL11 mutant during typical axenic culture. While bacteria cultured in the microdialysers are not forming biofilms per se, these results suggest that similar changes in the mycobacterial cell wall lipid composition occur in the confined space of a micro reactor as in a biofilm that result in a drug-tolerant phenotype. In the absence of MmpL11, M. smegmatis is unable to establish this drug-resistant state.
Unlike bactericidal antibiotics, bacteriostatic drugs ultimately depend upon the immune system for sterilization, and are therefore poor treatment options where the immune system is compromised [52]. Distinguishing between bacteriostatic and bactericidal action of antimicrobial drugs can be cumbersome using conventional drug susceptibility testing but the microdialyser can rapidly resolve this distinction for antimicrobial agents. Using the microdialyser process, we substituted the rifampicin-containing medium with drug-free medium without otherwise perturbing the cells in the non-growing cultures. Upon withdrawal of rifampicin via microdialysis, growth resumed in all the larger reactors (Fig 8), suggesting that the growthinhibitory effect of rifampicin was of the bacteriostatic kind. It is worth noting that some conventional studies have indicated that rifampicin is bactericidal to M. smegmatis cells at 32 μg/ml [30].

Conclusions
A physical cell culture system that mimics confinement of mycobacteria such as in a macrophage during infection enabled us to investigate confinement-induced drug tolerance in M. smegmatis cells. We uncovered an epigenetic rifampicin-resistance phenotype that was dependent on the size of the growth chamber: appearing when the mycobacteria were cultured in the small (200pL) growth chambers and disappearing in the bigger reactor volumes (500pL, 1200pL and 1700pL). The drug-tolerant phenotype was mediated in large part by efflux mechanisms, which were induced in confined growth environment. The confinement-induced drug tolerance observed in the microdialyser has similarities with the macrophage-induced tolerance that others have reported in intramacrophage mycobacterial species [18]. To the degree that confinement is a significant common factor for mycobacteria replicating in small microdialyser reactors or macrophages, macrophage-induced drug tolerance may be an effect of confined growth in addition to other macrophage specific mechanisms.
In vitro drug susceptibility tests performed using conventional cell culture systems, which are the basis for drug prescriptions, are notoriously poor predictors of treatment outcomes in human TB [53,54]. Our results suggest that the volume difference between the in vitro and in vivo (intramacrophage) growth compartments for M. tuberculosis, together with its implications for epigenetic drug resistance, may be a contributor to the apparent incongruity between drug susceptibility tests and treatment outcomes. Indeed, unlike the mycobacteria freshly inoculated into a conventional culture vessel, intramacrophage mycobacteria, which comprise a significant portion of TB infection [16], experience a space-confined growth environment that our experiments suggest can induce drug tolerance. Confinement-induced drug tolerance in M. tuberculosis may contribute to persistence in tuberculosis patients, where drugs that are rapidly potent in vitro require prolonged administration to achieve comparable effects. The microdialyser may thus provide an appropriate paradigm for research on therapeutic interventions aimed at rapidly neutralizing the drug-tolerant mycobacteria that currently prolong treatment in human TB.  Use of the microdialyser system to resolve the bacteriostatic/bactericidal credentials of antimicrobial drugs. The graphs in the left column represent a typical growth curves for each growth chamber size and those in the right column represent a set of growth curves for the same growth chamber volume. The cultures represented by the red curves were first cultured in drug-free medium for 15 hours. Using the microdialysis process, at 15 hours, rifampicin (350 μg/ml) was introduced to these cultures and then withdrawn at 150 hours. Although microbial growth in the bigger volume reactors-seven 500, eight 1200 and eight 1700pL-was effectively suppressed when rifampicin was applied, cell growth resumed when the drug was withdrawn, suggesting that rifampicin was bacteriostatic under these conditions. The cells in one of the eight 500pL maintained slow growth in the presence of the drug. Microbial growth in five of six small (200pL) growth chambers was only slightly decreased by the drug and suppressed in one of the reactors in which the drug was bacteriostatic. Control cultures are depicted by black curves demonstrating consistently positive growth in drug-free medium across all growth chamber sizes. (B) A system for distinguishing the bacteriostatic/bactericidal credentials of antimicrobial drugs. The X-axis represents the cell density reached by ΔFtsEX cells in the reactors shown above (Fig 8A) after culturing for 150 hours in medium with rifampicin (350 μg/ml) during the first phase of culture. The Y-axis represents the new cell density attained in the same reactors after 150 additional hours of culturing in medium without rifampicin. Accordingly, data points in quadrants 1, 2 and 3 represent reactors in which the cells are dead, static or resistant respectively. In general M. smegmatis ΔftsEX cells were resistant to rifampicin in the 200pL reactors and static in the rest.

Supporting Information
doi:10.1371/journal.pone.0136231.g008 Confinement-Induced Drug-Tolerance in Mycobacteria link valves are labelled (see Fig 1). With the link valve closed, the conditioning chamber of the middle microdialyser unit is filled with blue dye (representing fresh medium). Scale bar, 0.3mm. (C) Once the link valve is open, diffusive exchange between the growth and conditioning chambers occurs. (D) After a series of (typically six) microdialysis steps, the growth chamber fluid is completely replaced with the fluid introduced via the conditioning chamber.