ABSTRACT
During the dormant phase of tuberculosis, Mycobacterium tuberculosis persists in lung granulomas by residing in foamy macrophages (FM) that contain abundant lipid bodies (LB) in their cytoplasm, allowing bacilli to accumulate lipids as intracytoplasmic lipid inclusions (ILI). An experimental model of FM is presented where bone marrow-derived mouse macrophages are infected with M. avium and exposed to very-low-density lipoprotein (VLDL) as a lipid source. Quantitative analysis of detailed electron microscope observations showed the following results. (i) Macrophages became foamy, and mycobacteria formed ILI, for which host triacylglycerides, rather than cholesterol, was essential. (ii) Lipid transfer occurred via mycobacterium-induced fusion between LB and phagosomes. (iii) Mycobacteria showed a thinned cell wall and became elongated but did not divide. (iv) Upon removal of VLDL, LB and ILI declined within hours, and simultaneous resumption of mycobacterial division restored the number of mycobacteria to the same level as that found in untreated control macrophages. This showed that the presence of ILI resulted in a reversible block of division without causing a change in the mycobacterial replication rate. Fluctuation between ILI either partially or fully extending throughout the mycobacterial cytoplasm was suggestive of bacterial cell cycle events. We propose that VLDL-driven FM constitute a well-defined cellular system in which to study changed metabolic states of intracellular mycobacteria that may relate to persistence and reactivation of tuberculosis.
INTRODUCTION
Tuberculosis (TB) is one of the leading causes of mortality in the world with approximately 1.5 million deaths per year (1). It is estimated that one third of the world population is infected with the etiological agent of TB, Mycobacterium tuberculosis, but only 5 to 10% of infected individuals are at risk of developing active disease within a few years (2). Most individuals remain asymptomatic after the primary infection. In such individuals, bacilli may persist for decades in a dormant state until reactivation of bacilli leads to active disease (3).
Infection with M. tuberculosis follows a relatively well-defined sequence of events (reviewed in references 4 and 5). Infection is initiated when inhaled bacilli are phagocytosed by host alveolar macrophages. Once inside macrophages, bacilli reside and replicate in phagosomes that they prevent from maturing and, as a result, from fusing with lysosomes. In this manner, pathogenic mycobacteria avoid the acidic and cytolytic environment of a phagolysosome (reviewed in references 4, 6, and 7). During this phase, mycobacteria grow exponentially until the emergence of an acquired immune response takes effect (5). Concomitantly, and within weeks after M. tuberculosis infection, a granuloma is formed at the site of infection in the lungs. The granuloma is composed of a core of infected alveolar macrophages, surrounded by additional types of macrophages, viz., monocytes, multinucleated giant cells, epithelioid cells, and most notably, foamy macrophages (FM) (reviewed in references 4 and 5). The foamy aspect of these cells results from the accumulation of neutral lipids, typically triacylglycerides (TAG) and also esterified and nonesterified sterols within lipid bodies (LB), also called lipid droplets or lipid vacuoles (8–10). FM play a central role during the symptom-free, dormant phase of infection (11) until the host immune system is weakened and M. tuberculosis is reactivated to cause active disease.
FM have been found within human TB patients (12–14), in Schwann cells of leprosy patients (12, 15), in Mycobacterium avium-infected AIDS patients (16), as well as in chronic stages of M. tuberculosis infection in mice (17–19). Bacilli in viable tissue are located mainly in FM in both human and murine disease (13). In developing foci of necrosis, the bacilli are found in greater numbers associated with LB in FM (13).
Upon infection with M. tuberculosis, in vitro-grown human peripheral blood mononuclear cell (PBMC)-derived macrophages (HMDM) acquire a foam cell phenotype, without any other additives (20–22). This has never been observed in bone marrow-derived mouse macrophages (BMDM) infected with either Mycobacterium avium (6), Mycobacterium bovis BCG (C. de Chastellier, unpublished data), or M. tuberculosis (C. de Chastellier, unpublished data). This may be due to species or cell type differences in lipolytic activity in M. tuberculosis-infected cells (21). It may also depend on culture conditions and more specifically the lipoprotein, free lipid, or lipid precursor composition of the serum used to culture cells. HMDM are often cultured with human serum (HS), and the composition of the serum may vary extensively among donors. This is corroborated by the fact that uninfected and M. tuberculosis-infected HMDM, cultured in fetal bovine serum (FBS) display far less, if any, LB than those cultured in HS (C. Astarie-Dequeker and C. de Chastellier, unpublished data). Interestingly, uninfected Schwann cells display LB, and the number and size of LB increase after infection with Mycobacterium leprae (15). Infection, therefore, does not seem to be the only signal to trigger LB formation.
Several groups have reported that when M. tuberculosis or M. bovis BCG reside in different types of FM, they accumulate lipids in the form of intracytoplasmic lipid inclusions (ILI) (11, 23, 24). The presence of ILI, initially observed in in vitro-grown M. tuberculosis and in sputum (25, 26), the loss of affinity for acid-fast stain seemingly as a result of an altered cell wall (27), and arrest of bacterial multiplication (23) are thought to be the major hallmarks of dormant bacilli.
To gain insight into cellular and molecular events and decipher underlying mechanisms of mycobacterial disease, simplified and better defined experimental models and approaches have been developed with the aim of simulating the conditions for dormancy and regrowth in M. tuberculosis. Because lung granulomas present a hypoxic environment (28, 29), in vitro-growing mycobacteria (M. tuberculosis and M. bovis BCG) have been subjected to oxygen depletion (30). In this hypoxic environment, bacilli accumulate triacylglycerides within ILI and enter into a nonreplicative state; upon reexposure to oxygen, the pool of TAG within ILI is dramatically reduced, and bacilli resume growth (31, 32). Likewise, when M. tuberculosis-infected THP1 cells and human macrophages derived from blood monocytes (HMDM) are grown under conditions of hypoxia, they acquire a foam cell appearance, and the bacilli accumulate host-derived TAG in the form of ILI and stop replicating (23). To the extent that in vitro culturing of cells, with or without hypoxia, may lead to some cell death and the resulting release of fatty acids, the latter may serve as a source for TAG in both cellular LB and mycobacterial ILI. Likewise, M. tuberculosis, mostly in human macrophages, perturbs cellular lipid homeostasis, which results in the accumulation of LB in macrophages (20–22). A study performed with in vitro-grown human granulomas, obtained by infection of human peripheral blood monocytic cells with M. tuberculosis (33), has shown that the bacilli within granuloma FM enter into a nonreplicating state but replicate normally within nonfoamy macrophages in the same granuloma (11). In this model, M. tuberculosis-containing phagosomes and LB in infected foamy macrophages are initially located at random but then appear in close proximity. Bacilli are then found in the same organelle as LB-derived lipids, although fusion between phagosomes and LB has not been demonstrated. Subsequently, intraphagosomal mycobacteria themselves accumulate lipid in the form of ILI (11). In general, it appears that lipid-rich FM within human granulomas provide a source of nutrients and energy for persistent M. tuberculosis (11).
Whereas the in vitro-grown human granuloma model (11, 34) is a promising tool for studying granuloma formation, dormancy, and reactivation, its complexity causes some experimental limitations. For example, it is difficult to obtain preparations of either pure FM or pure nonfoamy macrophages and to undertake precise kinetic studies of cellular events at the subcellular level during transition of one to the other. Our goal was, therefore, to develop a simpler and better defined experimental model of FM that would allow easier manipulation of cellular events.
We chose to expose M. avium-infected macrophages to lipoprotein for uptake by receptor-mediated endocytosis. There are multiple reasons for using this experimental approach. BMDM were infected with M. avium, rather than with M. tuberculosis, because the former is easier to use and has been shown to simulate the intracellular behavior of its highly pathogenic relative (reviewed in references 6 and 7). As an opportunistic mycobacterium, M. avium causes disseminated disease, usually in persons with advanced HIV infection, and pulmonary disease in persons whose systemic immunity is intact (35). M. avium also causes the formation of granulomas upon infection of both mice and humans, and FM were observed in M. avium-infected AIDS patients (16). Lipoproteins, especially initially, could provide lipids for the development of FM that later are located close to the necrotic center (11) where the caseum contains large amounts of cholesterol ester, cholesterol, TAG, and phospholipid (20) from damaged/decaying cells. Lipids in lipoproteins that are internalized by receptor-mediated endocytosis will undergo hydrolysis in lysosomes and will provide fatty acids for the biosynthesis of TAG for LB. Lipoproteins thus could present a means for specifying the substrate(s) for the formation of LB and ILI. Finally, very-low-density lipoprotein (VLDL) was preferred to low-density lipoprotein (LDL) and high-density lipoprotein (HDL) because of its high TAG content (36). In this study, we showed that this experimental approach allowed observations on morphological and metabolic changes of mycobacteria in response to accumulation/hydrolysis of intracellular lipid.
MATERIALS AND METHODS
Reagents.Dulbecco's modified Eagle's medium (DMEM) and glutaraldehyde grade I (electron microscopy [EM] grade) were purchased from Sigma (St. Louis, MO, USA). Fetal bovine serum (FBS) was from Biowest (Nuaillé, France), and phosphate-buffered saline (PBS) was from GIBCO (distributed by Invitrogen, Villebon sur Yvette, France). Commercial lipoproteins (VLDL, LDL, and HDL) were from Calbiochem-Merck (Darmstadt, Germany), and osmium tetroxide and Spurr resin were from Electron Microscopy Sciences (distributed by Euromedex, Mundolsheim, France). Bovine serum albumin conjugated to gold particles (BSA-Au) was purchased from the Utrecht University School of Medicine (Utrecht, The Netherlands).
Mycobacterium avium.The transparent (Tr) colony variant TMC 724 (serovar 2), initially obtained from the Trudeau Mycobacterial Culture Collection, was prepared as described in previous work (37). Briefly, we expanded bacteria after the first passage after isolation from the liver of C57BL/6 mice infected 6 to 8 weeks previously. Bacteria used for experiments were always of the first passage grown on Middlebrook 7H10 agar plates, supplemented with 0.5% Tween 80, 0.2% glycerol, and 10% oleic acid-albumin-dextrose-catalase (OADC). Aliquots of bacterial suspensions were concentrated in Middlebrook 7H9 medium devoid of Tween 80 and stored at −80°C. When required, frozen samples were quickly thawed, vortexed, and adjusted to the desired titer in complete cell culture medium. More than 95% of the bacteria were morphologically intact and viable at this stage.
Generation of a GFP-expressing strain of M. avium.M. avium TMC 724 was grown at 37°C in Middlebrook 7H9 broth (Difco) supplemented with 0.05% Tween 80 and 10% ADC (Difco). When the bacterial culture reached an optical density at 600 nm of 0.4, glycine was added to a final concentration of 1.5%, and bacteria were incubated overnight for an additional 12 h. After centrifugation at 2,000 × g for 10 min, the cells were washed three times with water supplemented with 10% glycerol and 0.05% Tween 80 and then diluted in 10% glycerol and 0.05% Tween 80 at 1/100th of the starting culture volume. Three micrograms of plasmid DNA (pCG211) (38) was added to 200-μl aliquots of electrocompetent bacteria. Electroporation was performed in cuvettes with 0.2-cm-gap width (Molecular BioProducts [MBP]) at room temperature using a gene pulser (Bio-Rad) with settings of 2.5 kV, 25 μF, and 200 Ω (39). Following electroporation, 0.8 ml of 7H9 broth plus ADC and Tween 80 was added, and the suspension was transferred into sterile vent screw cap tissue culture flasks. After incubation at 37°C overnight, serial dilutions of bacteria were plated onto selective 7H11 agar and incubated for 14 days at 37°C. Selective agar plates contained either hygromycin B at 100 mg/ml or kanamycin at 50 mg/ml. Green fluorescent protein (GFP) expression was detected with an inverted fluorescence microscope with excitation at 490 nm and detection at 515 nm.
BMDM culture and phagocytic uptake.Bone marrow cells were isolated from the femurs of 6- to 8-week-old C57BL/6 female mice and seeded onto tissue culture dishes (Falcon; Becton, Dickinson Labware, Meylan, France) 35 mm in diameter (4 × 105 cells per dish) or on 24-well tissue culture plates (1 × 105 per well). The culture medium was DMEM with high glucose (1 g liter−1) and high carbonate (3.7 g liter−1) concentrations supplemented with 10% heat-inactivated FBS, 10% L-cell conditioned medium (a source of colony-stimulating factor 1 CSF-1), and 2 mM l-glutamine. Five days after seeding, the adherent cells were washed twice with DMEM and refed with complete medium. Medium was then renewed on day 6. No antibiotics were added. Particles were added to 7-day-old macrophage cultures as follows. (i) The cells were infected for 4 h at 37°C with M. avium at a multiplicity of infection (MOI) of 20 for EM studies and at a MOI of 10 for CFU counts, washed in four changes of ice-cold PBS to eliminate noningested bacteria, and further incubated in complete medium devoid of antibiotics. (ii) The cells were first infected with M. avium, and 6 days later, they were given native hydrophobic or carboxylated hydrophilic 1-μm-diameter latex beads for 45 min. The latex bead solutions were diluted 1,000-fold in complete medium to obtain adequate particle uptake. The cells were then washed with PBS as described above. For long-term cultures, the medium was changed twice a week.
Treatment with lipoproteins and chase after treatment.After active replication of M. avium for 5 to 7 days, infected cells were exposed to lipoproteins for the indicated times. Three different lipoproteins were used: VLDL, LDL, and HDL. In the specific case of VLDL, the volume was adjusted so as to expose cells (1 × 106 per dish) to 140 μg of TAG per ml of medium, unless otherwise indicated. In some instances, infected cells were exposed to both VLDL (140 μg/ml TAG) and 80 μg/ml tetrahydrolipstatin (THL) for 24 h.
Labeling of lysosomes with BSA-Au.To label lysosomes, M. avium-infected macrophages were rinsed with serum-free medium and incubated for 30 min at 37°C with colloidal gold particles of 10 nm in diameter conjugated with BSA-Au diluted in serum-free medium. The cells were washed three times with serum-free medium and further incubated in complete medium devoid of BSA-Au for 2 h at 37°C. BSA-Au-treated cells were then exposed to VLDL and THL for 24 as described above. Lipid-filled lysosomes were examined for the presence or absence of BSA-Au.
Evaluation of viable intracellular mycobacteria.At selected times after infection with M. avium, the medium was removed. The cells were washed twice with PBS and lysed with 0.1% Triton X-100 (in distilled water). The number of viable bacteria per well was determined by plating 10-fold dilutions of macrophage lysate on Middlebrook 7H10 agar. Colonies were counted after incubation at 37°C for 14 days. For each time point, counts were made from three different wells. The number of macrophages per well was determined at the same time points by counting the number of nuclei after cell lysis with 0.1% Triton X-100.
Processing for conventional electron microscopy.Cells were fixed for 1 h at room temperature with 2.5% glutaraldehyde in 0.1 M Na cacodylate buffer (pH 7.2) containing 0.1 M sucrose, 5 mM CaCl2, and 5 mM MgCl2, washed with complete cacodylate buffer, and postfixed for 1 h at room temperature with 1% osmium tetroxide in the same buffer devoid of sucrose. They were washed with buffer, scraped off the dishes, concentrated in 2% agar in cacodylate buffer, and treated for 1 h at room temperature with 1% uranyl acetate in Veronal buffer. Samples were dehydrated in a graded series of ethanol solutions and embedded in Spurr resin. Thin sections (70 nm thick) were stained with 1% uranyl acetate in distilled water and then with lead citrate.
Epifluorescence microscopy.Cells grown on glass coverslips were infected for 4 h with GFP-expressing M. avium. All of the following steps were carried out at room temperature. At selected time points postinfection, the cells were fixed for 30 min with 3% paraformaldehyde in PBS buffer (pH 7.4) and washed in several rinses of PBS. Coverslips were washed in distilled water and mounted in Moviol. Cells were observed with an epifluorescence microscope (Leica).
Quantification.At the time points indicated in the figures, 150 to 300 intraphagosomal mycobacteria per sample were examined by using an EM to score the percentage of each category of M. avium ILI profiles. Cells were examined at random, and care was taken to avoid serial sections.
RESULTS
Generation of FM by exposure to VLDL.When mycobacteria reside in phagosomes of foam cells, both the LB that give the cells their foamy appearance, as well as the ILI that persistent bacilli accumulate in, are made up of triacylglycerides (TAG) as the major lipid component (9, 40, 41). Because of the high TAG content of very-low-density lipoprotein (VLDL) (36), we considered it a reliable source of fatty acids for LB and ILI formation in infected macrophages. Bone marrow-derived mouse macrophages (BMDM) were infected with M. avium, and after active mycobacterial replication for 6 days, the cells were exposed to VLDL. The cells were then processed for electron microcopy (EM), and thin sections were examined using the EM for the presence or absence of LB. We opted for commercially available VLDL because its composition was more reliable than that of patient-derived VLDL. Depending on the batch, VLDL contained about 4 to 6 times more TAG than cholesterol. The most reproducible results, both in terms of LB content and preservation of macrophage ultrastructure and integrity (less than 5% cell lysis), were obtained by exposing 1 × 106 macrophages (either noninfected or M. avium infected) to cell culture medium containing small amounts of fetal bovine serum (2% instead of 10%) to which VLDL was added. The volume of VLDL was adjusted so as to expose cells from 140 to at most 210 μg of TAG per ml of medium. Exposure to higher concentrations of TAG led to loss of macrophage integrity and eventual lysis of 25% of the cell population within 24 to 48 h of treatment. With lower concentrations of TAG (70 μg/ml or less), LB were scarce. Accordingly, an amount of VLDL was applied in the following experiments to yield a concentration of 140 μg of TAG per ml of medium, unless indicated otherwise.
Without exposure to VLDL, macrophages were devoid of LB (Fig. 1A). In both noninfected and M. avium-infected macrophages, exposure to VLDL resulted in the formation of large numbers of LB, and 90% of the macrophage profiles became foamy in appearance (Fig. 1B). Due to its lipid content, the LB core appeared uniformly electron translucent. LB were spherical and about 0.3 to 0.6 μm wide. LB could not be mistaken for endocytic organelles, because the latter contain dense material and are limited by an electron-dense phospholipid bilayer membrane that is visible by using an EM. In contrast, the phospholipid monolayer that must be assumed to cover the LB outer surface could not be detected under the present EM conditions.
Exposure of BMDM to VLDL induces foam cell formation. BMDM were infected with M. avium. At day 6 p.i., cells were exposed to VLDL for 0 to 24 h, fixed, and processed for EM. (A) General view of a cell in normal medium devoid of VLDL. No LB are formed. (B) General view of a cell after a 24-h exposure to VLDL. The cell displays large amounts of LB as is typical of FM. Bars in panels A and B, 1 μm. (C) Time course of LB formation during exposure of uninfected or infected BMDM to VLDL. The relative abundance of LB was calculated as the percentage of LB-containing cell profiles × average number of LB per cell profile. The curves were drawn by eye. The values shown are accurate to about ±12% based on the evaluation of 60 to 100 independent cell profiles for each value.
At selected time points of exposure to VLDL (0 to 24 h), the amount of LB per cell profile was scored. LB started to form within hours after exposure to VLDL, and their relative number increased with time until steady state was reached after about 24 h (Fig. 1C). Infected cells displayed almost twice the number of LB as uninfected cells did (closed versus open symbols in Fig. 1C). This suggested that intracellular mycobacteria played a stimulatory role for LB formation by their host cell, in agreement with previous studies (11, 21, 42). However, in view of the 4-h delay of enhanced increase of LB abundance in infected cells (cf. closed symbols in Fig. 1C), it seemed that such an effect only set in after the host cells had accumulated a threshold abundance of LB (as for uninfected macrophages; cf. open symbols in Fig. 1C). This observation might suggest that mycobacteria required time to affect the metabolism of the host cell.
This VLDL-induced model of FM was then used to examine changes in mycobacterial morphology and growth with the aim of determining whether mycobacteria acquired the major characteristics of persistent bacilli.
Mycobacteria in VLDL-driven FM accumulate lipids in the form of ILI.It is well-known that persistent mycobacteria accumulate large amounts of TAG in the form of ILI and use this as a source of carbon and energy for maintenance and eventual regrowth (31, 32, 43–45). To characterize the process of ILI formation, M. avium-infected macrophages were exposed to VLDL for 0 to 72 h, and profiles of intracellular mycobacteria were examined by EM for the presence and extent of ILI formation in both VLDL-treated and untreated macrophages. Mycobacteria displayed different profiles in terms of ILI size, and they were divided into 4 categories according to their ILI size (Fig. 2A to D), viz., ILI− profiles with no ILI (Fig. 2A), ILI+1 for profiles displaying a few small ILI 0.1 μm in width at most (Fig. 2B), ILI+2 for profiles displaying several ILI approximately 0.2 to 0.3 μm in width (Fig. 2C), and ILI+3 for profiles displaying several ILI approximately 0.4 to 0.5 μm in width and occupying most of the mycobacterial cytoplasm (Fig. 2D). The relative abundance of each type of ILI profile was scored for the indicated times of exposure to VLDL (Fig. 2E). In the absence of VLDL treatment, 95% of the M. avium profiles were always of the ILI− or ILI+1 category. Upon exposure to VLDL, mycobacteria with ILI+2 profiles started to appear immediately but declined after a lag of about 2 h concurrent with the appearance of ILI+3 profiles displaying the most extensive lipid accumulation. This suggested that mycobacteria with ILI+2 subsequently developed into mycobacteria with ILI+3. The nonzero steady-state value for ILI+2 at about 30% implied that mycobacteria with ILI+3 could revert back to ILI+2. Evidence for such a dynamic relationship between the ILI+2 and ILI+3 states was even more pronounced when looking at their respective appearances as a function of LB formation during their host cells' exposure to VLDL (Fig. 2F). Lower levels of LB in cells resulted mainly in ILI+2 profiles, whereas the abundance of ILI+3 increased only when sufficient LB were available (Fig. 2F). Sufficient LB seemed to be a rate-limiting factor for the transition of ILI+2 to ILI+3, but not a limiting factor for the formation of ILI+2 themselves.
Mycobacteria residing in VLDL-induced FM accumulate lipid in the form of ILI. M. avium-infected cells were exposed to VLDL for 0 to 72 h, fixed, and processed for EM. Bacterial profiles were divided into 4 different categories in terms of the presence and size of ILI (ILI−, ILI+1, ILI+2, and ILI+3). (A) ILI− is no ILI. (B) ILI+1 is less than 5 small ILI up to 0.1 μm in width. (C) ILI+2 is several ILI 0.2 to 0.3 μm in width. (D) ILI+3 is several ILI 0.4 to 0.5 μm in width and extending across the full width of the M. avium cytoplasm. Bars in panels A to D, 0.5 μm. (E) Time course of ILI formation during a 72-h period of exposure to VLDL. The combined contribution of ILI+2 and ILI+3 profiles was complementary (by way of percentage calculation) to the combined abundance (not shown) of ILI− and ILI+1 profiles. Error bars indicate the standard deviations (SD) based on the results of 2 to 4 independent experiments. For each experiment, 150 to 300 bacterial profiles were examined per time point. The curves were drawn by eye. (F) Dependence of ILI formation on the relative abundance of LB as obtained by plotting the ILI values from panel E as a function of the LB values in Fig. 1C (infected cells) at the corresponding time points. The curves were drawn by eye.
TAG, but not cholesterol, is essential for ILI formation.Foam cell LB contain TAG and esterified and nonesterified cholesterol (reviewed in references 46 and 47), but it has not been ascertained which of these compounds is required for ILI formation. It has been reported, however, that acquisition of host cholesterol through the mycobacterial Mce4 transporter system is critically linked to intracellular persistence of M. tuberculosis (48). To address this issue, infected macrophages were exposed for 24 h to different concentrations of three lipoproteins, VLDL, low-density lipoprotein (LDL), and high-density lipoprotein (HDL). VLDL contains 4 to 6 times more TAG than cholesterol, LDL contains 4 to 6 times more cholesterol than TAG, and HDL contains only minute amounts of both TAG and cholesterol. Formation of ILI+2 and ILI+3 profiles was monitored as based on the criteria defined in the legend for Fig. 2A to D. After exposure to HDL, the mycobacterial profiles belonged exclusively to the ILI− and ILI+1 categories (Fig. 3A). The formation of ILI+3 depended strictly on the presence of sufficient TAG, such that below about 70 μg/ml, no ILI+3 were observed (Fig. 3B). In comparison, ILI+2 already formed at TAG concentrations as low as 28 μg/ml and reached a maximum at about 70 μg/ml (Fig. 3B). At higher TAG concentrations where ILI+3 started to become abundant, the steady-state levels of ILI+2 declined somewhat, suggestive of the dynamic relationship between these two ILI states, as mentioned above in “Mycobacteria in VLDL-driven FM accumulate lipids in the form of ILI.” In view of the uncertainty regarding the roles of different LB-contained lipids, it was found that cholesterol could not compensate for insufficient TAG concentrations (Fig. 3C). As long as the TAG concentration remained below about 70 μg/ml, even cholesterol concentrations up to 128 μg/ml did not lead to the formation of any observable ILI+3 (Fig. 3C and D). These results showed that TAG, and not cholesterol, was essential for ILI+3 formation. However, because VLDL contains small amounts of cholesterol, the possibility that a minimum of cholesterol may, nevertheless, be required for ILI+3 formation could not be ruled out.
TAG, and not cholesterol, is essential for ILI formation. At day 6 p.i. with M. avium, BMDM were exposed for 24 h to different types and amounts of lipoprotein, resulting in different TAG concentrations: HDL, 2 μg/ml; native LDL, 7 to 28 μg/ml; VLDL, 70 to 215 μg/ml. The cells were then processed for EM and analyzed for mycobacterial ILI formation. (A) Treatment with HDL containing 2 μg/ml TAG and 6 μg/ml Ch. The profiles are ILI+1 or ILI− as for untreated cells. (B) Dependence of steady-state levels of ILI+3 and ILI+2 profiles on the TAG concentration. Error bars indicate the SD based on the results of 2 to 4 independent experiments. For each experiment, 150 to 300 bacterial profiles were examined per time point. (C) Infected cells were exposed to LDL containing low concentrations of TAG and either low (32 μg/ml) or high (128 μg/ml) concentrations of cholesterol (Ch). Error bars indicate the variation between 2 independent experiments. For each experiment, 150 to 300 bacterial profiles were examined per time point. (D) Morphological appearance of an ILI+2 mycobacterium after treatment with LDL containing 28 μg/ml TAG and 128 μg/ml Ch. Note the presence of Ch crystals. Bars in panels A and D, 0.5 μm.
Mycobacterial ILI formation depends on lipid processing by the host macrophage.The dependence of ILI formation on the presence of LB in the host (Fig. 2F) and the finding that TAG provided the bulk of lipid for ILI formation (Fig. 3) confirmed that TAG in ILI were derived from the host cell (21, 23). Therefore, we determined to what extent the mobilization of TAG from VLDL required processing by the host cell, in particular the action of host cell lipases for the release of fatty acids.
The first experiment served to determine whether M. avium can access TAG directly from VLDL without the involvement of the host macrophage. Mycobacteria were cultured in Middlebrook 7H9 medium for 4 days to allow for normal mycobacterial growth. VLDL was then added to the culture medium for 24 h in such a way that the concentration of VLDL was the same as for growth in host macrophages. Untreated M. avium was grown in parallel to serve as a control. In the presence of VLDL, in vitro-growing M. avium displayed the same morphology as M. avium grown in the absence of VLDL; in particular, there was no ILI formation (Fig. 4A versus B). These results showed that surface-exposed mycobacterial lipases were unable to directly process TAG and/or derivatives enclosed in the VLDL core, and as a result, fatty acids were not mobilized for import and ILI formation in the mycobacterial cytoplasm.
Fatty acids for LB and ILI formation are provided through TAG (from VLDL) breakdown in host lysosomes. (A and B) M. avium cultured in vitro in Middlebrook 7H9 medium for 4 days was exposed to VLDL for 24 h. Bacteria were processed for EM and examined for the presence or absence of ILI and division septa. Bacteria cultured with VLDL (A) or without VLDL (B) were devoid of ILI and displayed a division septum (arrow). (C and D) BMDM were infected with M. avium. Six days later, cells were first exposed to BSA-Au to label host lysosomes (Ly) (C) or not exposed to BSA-Au (D) before treatment for 24 h with VLDL in the presence of tetrahydrolipstatin (THL), a cellular and mycobacterial lipase inhibitor. The cells were then processed for EM and examined for LB and ILI formation. LB were scarce. VLDL accumulated in lysosomes that were easily recognizable by the presence of BSA-Au (arrow in panel C) or their dense material (arrows in panel D). ILI remained small (0.1 to 0.2 μm in width), and mycobacteria kept their normal length. Bars, 0.5 μm (A, B, and C) and 1 μm (D).
The requirement of host lipase action was tested by involving the antiobesity drug, tetrahydrolipstatin (THL), which inhibits human digestive lipases within the gastrointestinal tract (49) and a variety of other serine hydrolases (50–52), including mycobacterial LipY responsible for hydrolysis of ILI-containing TAG during the persistence and reactivation phases (43). When infected macrophages were exposed simultaneously to VLDL and THL, endocytosis of VLDL and transfer to lysosomes was not affected. However, TAG and derivatives in VLDL were not degraded as usual. They could easily be recognized by their electron transparency and were seen to fill the entire lumen of lysosomes (Fig. 4C and D), leaving only small spaces with the usual electron-dense lysosomal contents (Fig. 4D) and BSA-gold added to cells for lysosome identification (Fig. 4C). These data suggested that lysosomal TAG hydrolases were inhibited by THL. As a result, they were unable to break TAG down into di- and monoglycerides and fatty acids for cell reprocessing into TAG and accumulation in LB. In the presence of THL, very few LB were observed (about 1.6 LB per cell profile versus 15.2 in cells exposed to VLDL only). As expected, in the absence of LB formation, ILI+3 mycobacterial profiles reached only about 8% of the amount generated by cell exposure to VLDL without THL. ILI+3 profiles amounted to 3.7% in the presence of THL versus 43.5% in the absence of THL (average of 3 independent experiments). These results indicated that mycobacterial lipases were unable to process TAG in the VLDL core directly and that VLDL-derived TAG first had to be processed by lysosomal lipases.
Host lipids are made accessible to mycobacteria through fusion of macrophage LB with mycobacterium-containing phagosomes.So far, it has not been established how intraphagosomal mycobacteria acquire TAG from LB. It is not known whether TAG-rich LB are engulfed by phagosomes by a process reminiscent of autophagy or whether they fuse directly with mycobacterium-containing phagosomes (11). Mycobacteria normally reside in phagosomes that do not mature and therefore do not fuse with lysosomes to which endocytosed VLDL will be delivered for degradation. EM analysis of infected cells after a 4- to 24-h exposure to VLDL showed that LB were initially distributed randomly throughout the host cytoplasm (Fig. 1B), before establishing an intimate contact with phagosomes that led to a deformation of the phagosome membrane (Fig. 5A) and fusion of the two structures (Fig. 5B and C). Lipid seeped into the phagosome between its membrane and the surface of the mycobacterial cell wall (Fig. 5B and D) where TAG could become accessible for breakdown into fatty acids by mycobacterial lipases located at the outermost cell wall layer (53, 54). After discharge of their lipid contents, LB became completely electron translucent (Fig. 5E).
TAG is made accessible for ILI formation by fusion of host LB with M. avium-containing phagosomes. M. avium-infected BMDM were exposed to VLDL and processed for EM. (A) M. avium-containing phagosome in direct contact with host LB, showing deformation of the phagosome membrane into the LB (arrow). (B) Lipid from LB seeps into the phagosome (arrow). (C) Fusion between a LB and an M. avium-containing phagosome (arrows). (D) Discharge of host lipids into a phagosome. The lipids are in contact with the bacterial surface (arrowhead). (E) LB that have fused with an M. avium-containing phagosome become electron translucent (eLB), suggesting that they have discharged their lipid content into the phagosome. (F) BMDM were given hydrophobic or hydrophilic latex beads of 1 μm in size before being exposed to VLDL. A LB is seen close to a phagosome containing a hydrophilic latex bead [Ph(La)]. No deformations of the closely apposing phagosome membrane (arrows), and no phagosome-LB fusions were observed. Bars in panels A to F, 0.5 μm.
Mycobacteria survive in host macrophages by residing in immature phagosomes, which they prevent from maturing and subsequently fusing with lysosomes (reviewed in references 6 and 7). To maintain this maturation block, an all-around close apposition between the bacterial surface and the surrounding phagosome membrane must be maintained, a geometry that is obstructed when multiple mycobacteria reside in the same phagosome (63). In this case, maturation and fusion with lysosomes cannot be prevented, and mycobacteria end up in a phagolysosome. LB fused with both nonmatured phagosomes with a single bacterium and with phagolysosomes with one or several bacteria. M. avium was able to accumulate lipids in ILI whether it resided in a nonmatured phagosome or in a phagolysosome. Of interest for the latter case was the observation that fusion between LB and phagolysosomes seemed to occur only at sites where the phagosome membrane was closely apposed to the mycobacterial surface. Furthermore, LB interacted with no other organelles, including those of the endocytic pathway. None of the above events were observed when phagosomes contained other particles such as hydrophobic or hydrophilic latex beads (Fig. 5F). These observations suggested that contact, deformation, and fusion between LB and phagosomes happened under mycobacterial control.
Mycobacterial division is arrested in VLDL-driven foamy macrophages.Arrest of mycobacterial replication is considered to be a major characteristic of persistent bacilli (11, 23). Since this conclusion is based on CFU counts exclusively, which give information only on mycobacterial division, not on replication, it was of special interest to investigate whether mycobacterial division or replication was affected when M. avium formed ILI in VLDL-driven foamy macrophages. BMDM were infected with M. avium at an MOI of 10 for 5 days. After active replication of bacilli, infected cells were exposed to VLDL for 0 to 5 days. At selected time points during the treatment, cells were lysed, and the number of CFU were determined. The number of viable bacilli in untreated macrophages (control cells), cultured in parallel, was determined from the onset of infection (day 0) up to day 12 postinfection (p.i.). In untreated cells, the M. avium doubling time (T2) was 1.7 days (Fig. 6A), as also observed in previous studies (55, 56). In contrast, the CFU count remained constant during exposure to VLDL (Fig. 6A). Only a small increase at about 20% of the rate of the control values was noticeable after 5 days (dotted line in Fig. 6A). These results indicated that about 80% of M. avium stopped to divide when their host cells were exposed to VLDL.
Division arrest and elongation of M. avium in VLDL-induced FM. BMDM were infected with M. avium and exposed to VLDL for 1 to 5 days to observe the effects on mycobacterial growth. (A) The number of CFU was determined at the indicated times p.i., either in untreated control cells (closed symbols) or in cells treated with VLDL (open symbols). The data for control cells indicated a doubling time (T2) of 1.7 ± 0.07 days. In the case of VLDL-treated cells, CFU counts remained constant (dashed line; with the possibility of an increase at about 20% of the control rate [dotted line]). Different symbols correspond to data from 3 independent experiments. (B, C, and D) EM observations show the normal length of a mycobacterium (approximately 1 μm) without VLDL treatment (B), the appearance of M. avium after a 1-day (C) or 5-day (D) exposure to VLDL. Note the elongation of M. avium resulting from exposure to VLDL and the absence of a septum. Bars in panels B, C, and D, 0.5 μm.
Arrest of mycobacterial division does not necessarily imply a cessation of metabolism and replication. Measurements of CFU reflected only the number of mycobacteria, not their morphology. Therefore, their appearance was examined by epifluorescence microscopy (not shown) and EM (Fig. 6B, C, and D). A GFP-expressing strain was used. During replication within normal macrophages, this strain had the same growth characteristics and morphological features as the non-GFP-expressing strain from which it was constructed. In particular, GFP-expressing mycobacteria had the same length (0.8 to 1 μm long), doubling time (around 1.7 days), and cell wall appearance, including a thick (approximately 80-nm-wide) electron-translucent zone. After a 24-h exposure to VLDL, more than 90% of the mycobacteria were longer than those observed in untreated macrophages (Fig. 6C versus B). Mycobacteria continued to elongate, and after 5 days of treatment with VLDL, they were about eight times longer than those observed in control cells (Fig. 6D versus B), reflecting an unchanged rate of mycobacterial growth (8 = 23, or 3 doublings in length in 5 days, i.e., T2 = 1.7 days). The rate of elongation indicated that mycobacteria remained metabolically active at the same level as under lipid-free conditions.
Mycobacteria were then examined for the presence or absence of a division septum. Between 100 and 200 longitudinal sections of M. avium were examined at selected time points of treatment (0 to 3 days) in three separate experiments. Whereas 6 to 10% of the longitudinal mycobacteria within control macrophages displayed a division septum, none of the bacteria displaying an ILI+3 or ILI+2 profile had a septum after exposure of infected cells to VLDL for 24 h or longer. These results indicate that, while mycobacteria in VLDL-driven FM accumulated large amounts of lipids in the form of ILI, they did not divide but continued to elongate (Fig. 6C and D). Interestingly, the cell wall electron-translucent zone (ETZ) of ILI+3 mycobacteria had become noticeably thin after a 24-h exposure to VLDL (Fig. 2D versus A). This could be due to modifications in the synthesis of mycolic acids, one of the major ETZ constituents, as has been observed in persistent bacilli (27).
Removal of VLDL leads to a rapid decline of LB in host macrophages, the concurrent disappearance of ILI from M. avium, and the immediate resumption of mycobacterial division.In about 10% of the cases of latent TB, M. tuberculosis is reactivated and causes acute disease. It was of interest, therefore, to investigate whether the effects of VLDL, viz., ILI formation and mycobacterial division arrest, could be reversed by its removal. After exposure of infected macrophages to VLDL for 1 to 5 days, the cells were washed and reincubated in fresh medium without lipoprotein. Within 24 h after removal of VLDL, infected and uninfected macrophages lost their LB (Fig. 7A) and reverted to their normal appearance (as in Fig. 1A). The morphological appearance of M. avium was examined by epifluorescence microscopy (data not shown) and EM. Regardless of the duration of the preceding exposure to VLDL, all mycobacteria resumed their normal length within 24 h to at most 48 h of chase in VLDL-free medium (Fig. 7B, panel c, versus Fig. 2A). It is noteworthy that the cell wall ETZ resumed its normal thickness within 24 to 48 h of chase (Fig. 7B, panels b and c versus panel a).
Removal of externally added VLDL induces a rapid decline of macrophage LB and mycobacterial ILI. At day 6 p.i., BMDM were exposed to VLDL for 24 h and then reincubated in VLDL-free culture medium for 0 to 3 days. At selected time points, the effects of VLDL removal on the decline of LB in macrophages and ILI in mycobacteria were examined. (A) LB decline in uninfected and infected macrophages. The higher starting level for infected macrophages resulted from their twofold-higher steady-state level of LB formation (cf. Fig. 1C). Data are compatible with a first-order process with a half-life (T1/2) as indicated. The values shown are accurate to about ±12% based on the evaluation of 60 to 100 independent cell profiles for each value. (B) Decline in mycobacterial ILI size from ILI+3 to ILI+1/− over a 2-day period of chase. (a) VLDL for 24 h and no chase; (b) VLDL for 24 h, followed by a 1-day chase; (c) VLDL for 24 h, followed by a 2-day chase. Mycobacteria have fully recovered their normal length and their electron-translucent cell wall layer (panel c versus a). (d and e) Mycobacterial cross sections illustrate the gradual disappearance of ILI from full size ILI+3 (numbered 1 in panel d) through progressive stages of ILI shrinkage (numbered 2 and 3 in panel d) or fragmentation of ILI into smaller lipid droplets (panel e). Bars in panels a to e, 0.5 μm. (C) Time course of the decline in ILI+3 and ILI+2 profiles upon removal of VLDL. The combined contribution of these two types of ILI profiles is complementary (by way of percentage calculation) to the combined contribution by ILI− and ILI+1 profiles (not shown). A total of 150 to 300 bacterial profiles were examined per time point. The lines were drawn by eye. (D) Decline of ILI profiles when expressed in relation to decreasing LB levels. The ILI+3 and ILI+2 values from panel C were plotted as a function of the LB values from panel A (for infected cells) at the corresponding time points. Logarithmic scales allow for sufficient visual resolution of the difference between low LB values at later times. The lines were drawn by eye.
Because ILI formation depended on the presence of LB in the host macrophage (Fig. 2F), the fate of mycobacterial ILI after removal of VLDL was determined by EM. At the onset of the chase period, 50% of the bacteria were ILI+3 (Fig. 7B, panel a). By 24 h of chase, ILI were considerably reduced in size, being either ILI+2 (Fig. 7B, panel b) or ILI+1 (Fig. 7B, panel c). By 48 h, all the bacterial profiles were ILI+1 or devoid of ILI. Cross sections of bacteria showed disappearance of ILI from full size ILI+3 (numbered 1 in Fig. 7B, panel d) through progressive stages of ILI shrinkage (numbered 2 and 3 in Fig. 5, panel d) or fragmentation (Fig. 7B, panel e). The rates of ILI+3 disappearance are shown in Fig. 7C. During this decline, the relative abundance of ILI+2 profiles initially increased from about 35% to over 50% after 8 h. This indicated that the disappearance of ILI+3 resulted from their transition to ILI+2 profiles. At 48 h after removal of VLDL (not shown in Fig. 7C), ILI+2 profiles had declined to a negligible fraction, and no ILI+3 profiles remained at this stage. To demonstrate the dependence of ILI on the presence of LB more directly, Fig. 7D shows the relative abundance of ILI profiles as in Fig. 7C as a function of LB abundance as in Fig. 7A. The linear dependence of ILI+3, but not of ILI+2, on LB abundance indicated that the presence of LB was a rate-limiting factor for the formation and maintenance of ILI+3.
The formation of ILI was associated with mycobacterial elongation, which implied division arrest. It was therefore of interest to determine whether this block in mycobacterial division was reversed when ILI disappeared as a result of VLDL removal. The data in Fig. 8A showed that this was indeed the case. After CFU counts remained stationary in the presence of VLDL for 1, 3, or 5 days, removal of VLDL led to a recovery of mycobacterial number. Not only was this recovery rapid in that it occurred in a fraction of a generation time but it was also complete to the extent that mycobacterial numbers were restored to exactly the level of their untreated counterparts. Septum formation was seen to occur after the transition of ILI+3 to ILI+2 (Fig. 8C [chase for 4 h] versus B [control]).
Rapid reversal of division arrest upon removal of VLDL. BMDM infected with M. avium were exposed to VLDL for 1 to 5 days and then reincubated in VLDL-free culture medium for an additional 1 to 4 days. (A) The number of CFU was determined at the indicated times p.i. either for untreated control cells (closed symbols), during cell exposure to VLDL (open symbols in gray-shaded areas), or during reincubation in VLDL-free medium (open symbols outside gray-shaded area). The data for control cells indicated a mycobacterial doubling time (T2) of about 1.8 ± 0.2 days. During exposure to VLDL, the CFU counts did not change. Upon removal of VLDL, the CFU counts increased rapidly to the value for the cells that had not been treated with VLDL. (B) ILI− profile showing a division septum (arrows) in an untreated cell. (C) Septum formation (arrows) in an ILI+2 profile 4 h after removal of VLDL. Bars in panels B and C, 0.5 μm.
Mycobacterial division arrest correlates with the presence of ILI+3.In order to examine a possible relationship between the presence of ILI and the inability of M. avium to divide, both the recovery of CFU counts and the time course of ILI decline were measured during the same experiment for the first 24 h of chase after VLDL removal. The time course of CFU counts showed a rapid and ultimately complete recovery of mycobacterial division when VLDL was removed (Fig. 9A). The approximate time course for the increasing CFU values is shown by the linear solid line in Fig. 9A. By interpolation along this line, additional CFU values could be constructed for intermediate time points that corresponded to times for which ILI values were measured. For CFU recovery, these CFU values were expressed as the percentage to which the linear line had caught up with the CFU level of the exponential growth curve of M. avium in the untreated cells (right-hand y axes in Fig. 9A and B and accentuated curve labeled “cfu” in Fig. 9B). CFU recovery values are shown in relation to the relative abundance of ILI+1 or ILI− (ILI+1/−), ILI+2, and ILI+3, respectively (Fig. 9C, D, and E).
Relationship between the ability of mycobacteria to divide and their ILI profile. BMDM were infected with M. avium, exposed to VLDL for 1 day, and then reincubated in VLDL-free culture medium for an additional 24 h, during which both the relative abundance of ILI profiles and CFU counts were measured. (A) Arrest and resumption of mycobacterial division observed via CFU counts (open symbols). Data are displayed on a linear scale, which shows the exponential characteristic of the control growth curve (closed symbols and fitted dashed-line curve with doubling time [T2] of 1.4 days). Values for the recovery of CFU counts at 0- to 24-h time points after removal of VLDL (right-hand y axis; not linear, ticks indicating 0, 25, 50, 75, and 100%) were obtained by interpolation along the linear solid line and expressed as a percentage to which this line had caught up with the values of the control growth curve (dashed line with closed symbols). (B) The values for CFU recovery as determined in panel A (right-hand y axis) are shown for the time points at which the relative abundances of ILI profiles were measured (left-hand y axis). (C to E) The CFU recovery values were related to the relative abundances of the various ILI profiles shown in panel B. The curves were drawn by eye.
Major CFU recovery occurred between 0 and 8 h, without a significant change in the values of ILI+1/− profiles, which remained around 20% (Fig. 9C), and also without a correlation to the levels of ILI+2 profiles (Fig. 9D). Only in the case of ILI+3 profiles did their gradual decline correlate with the increase of CFU recovery (Fig. 9E). This suggested that the block in mycobacterial division depended on the presence of ILI+3, but not on that of ILI+2 and ILI+1/−.
DISCUSSION
Under physiological conditions, neither monocytes nor lipoproteins will be found in the interstitium where a granuloma develops. Monocyte recruitment at such an inflammatory site may be accompanied by increased vascular permeability that would allow lipoproteins to leave the intravascular compartment, whereafter chemical modification enhances their uptake through scavenger receptors on macrophages. Necrotic cells could also contribute phospholipids and neutral lipids (20). Several experimental approaches have been developed to study the process of lipid mobilization for FM formation in granulomas (18, 21). These studies have shown that mycobacteria affect lipid metabolism in the host in ways that will promote LB formation in FM.
In the present study, lipoprotein (VLDL) was used as an experimental tool to generate FM, since it is known that such cells play a major role in the persistence and reactivation phases of tuberculosis. This model may thus apply to mycobacteria residing in macrophages, located in close vicinity to the necrotic center, that are known to acquire a foam cell appearance (11).
Our data showed that exposing M. avium-infected murine macrophages (BMDM) to TAG-rich VLDL, rather than to sterol-rich LDL, induced mycobacterial accumulation of massive amounts of lipids in the form of ILI. Because ILI formation is a major characteristic of persistent bacilli, that has been well established in several experimental models of FM or granuloma FM (11, 23, 24), the present experimental model, therefore, provided a useful tool to characterize events of ILI formation in the lipid-rich environment of a FM. Although a recent study has used oleate as a lipid source (21), the advantage of using VLDL instead was that in vitro-growing and intramacrophagic mycobacteria cannot gain sufficiently fast access to fatty acids to form ILI because of the low rate of fatty acid diffusion through membranes (57).
We believe that the strength of our experimental system resides in its ability to generate defined conditions for triggering the formation or removal of LB, which induces ILI formation or consumption, respectively, and in turn, leads to the reversible arrest of mycobacterial division without affecting bacterial growth. The use of EM approaches combined with determining the number of CFU has allowed us to demonstrate fusion of LB with mycobacterium-containing phagosomes as the mechanism for transfer of host TAG to phagosomes and to show an alteration of the mycobacterial cell wall by thinning out of the electron-translucent outermost layer. The latter might result from an altered mycolic acid synthesis that has been proposed as the reason for loss of acid fastness and resistance to antibiotics (27). Furthermore, the present approach has allowed us to qualitatively demonstrate (i) incremental stages of ILI formation and (ii) the presence or absence of septum formation in elongated cells and (iii) to uncover the formerly ignored difference between CFU counts and bacterial growth. Finally, observations were sufficiently reliable and reproducible for quantitative analyses that allowed us to characterize the relationship between fluctuations of ILI status and the arrest of mycobacterial division.
trans-Membrane action by mycobacteria is required for their acquisition of TAG from lipid bodies in the host cell.In order to gain access to TAG that is stored in lipid bodies of foamy macrophages, mycobacteria in host cell phagosomes must overcome partitions in the form of phospholipid layers. FM store TAG and/or cholesterol esters in the form of LB when cells are exposed to external sources of fat. When this source consists of VLDL, as in the present case, it is endocytosed and targeted to lysosomes for lipid hydrolysis to produce fatty acids, which would then be available for esterification to TAG and LB formation. The current models of LB biogenesis have been reviewed extensively (46, 47, 58). The prevailing model suggests that host LB originate at the endoplasmic reticulum (ER), where fatty acids are resynthesized to form TAG and sterol esters. These compounds accumulate in the hydrophobic space between the leaflets of the ER bilayer membrane, where they are subsequently engulfed by phospholipids of the cytosolic leaflet and pinched off the ER membrane into the cytoplasm. This process results in LB that are surrounded by a single layer of phospholipids. Several processes have been proposed for how invading pathogens can gain access to lipid bodies (reviewed in reference 47). Of these processes, direct fusion of the mycobacterium-containing phagosome membrane with lipid bodies has been considered an unlikely process, because lipid bodies are surrounded by a monolayer of phospholipids, whereas the phagosome membrane consists of a phospholipid bilayer. The present EM observations, however, provided visual evidence for direct fusion with host LB as a mechanism of host lipid release into mycobacterium-containing phagosomes. Such a process can be envisaged as the reverse of LB release from the ER into the cytoplasm as mentioned above. The monolayer phospholipid leaflet covering the LB could merge into the outer leaflet of the bilayer phagosome membrane. This would leave the lipid contents in the hydrophobic interior of the bilayer, where it remains separated from the mycobacterial surface by the inner leaflet. Because this remaining phospholipid barrier would present a different molecular arrangement than the otherwise intact phagosome bilayer membrane, it might be vulnerable to attack by mycobacterial lipolytic enzymes. There are 24 known lipid/ester hydrolases of M. tuberculosis (reviewed in reference 59), of which two in particular are known to degrade phospholipids, viz., phospholipase C (60) and Rv3452, a phospholipase A belonging to the cutinase family (61, 62).
It is important to note that lipid bodies were not observed to fuse with phagosomes that enclosed other types of particles, such as hydrophobic or hydrophilic inert latex particles, and also did not fuse with any of the different membrane compartments of the endocytic pathway. This indicated that the initial targeting and subsequent fusion process depended on specific effects caused by mycobacteria from within the phagosome across both bilayer membrane leaflets. Our previous work has shown that a prerequisite for mycobacteria to reside in nonmaturing phagosomes is that they maintain the phagosome membrane in close apposition to their surface all around (56, 63–65). The molecular basis for this close interaction is not known, but could provide the means to initiate fusion with LB and subsequently destabilize the relevant phospholipid layers and expose them to lipolytic attack. This might explain the observed deformation of the phagosome membrane where mycobacterium-containing phagosomes were in close contact with host lipid bodies. Further evidence for a trans-bilayer action of mycobacteria through the phagosome membrane came from observations where LB fused with phagosomes that contained multiple mycobacteria and for which the phagosome membrane was partially prevented from being closely apposed to the mycobacterial surface (e.g., where the membrane spans the region between two adjacent bacteria [65]). In such cases, fusion was observed only at sites where a close apposition could nevertheless be maintained.
Mycobacteria mobilize TAG from lipid bodies to form ILI.A previous study has shown that M. tuberculosis and M. bovis BCG in FM import fatty acids derived from TAG in the host to resynthesize TAG that is then stored as ILI (23). This result was reproduced in the present system of VLDL-driven FM. The formation and maintenance of ILI were strictly dependent on the presence of LB in the host cell, and once LB declined upon withdrawal of external VLDL, ILI in mycobacteria disappeared as well. TAG is the bulk lipid in ILI (31, 43, 44), and we showed that ILI formation strictly depended on sufficient amounts of TAG in the host cell. Increasing amounts of cholesterol could not compensate for limited availability of TAG. This may seem to contradict the work by Pandey and Sassetti (48), which shows that cholesterol is required for mycobacterial persistence, but was not aimed at analyzing ILI formation. Since VLDL is not free of cholesterol, we could not rule out the possibility that a minimum amount of cholesterol was required for ILI formation.
It is unlikely that intact TAG molecules from LB can transit the mycobacterial cell wall for ILI formation. At best, this would be an extremely slow process. Instead, acquisition of TAG for ILI formation can be expected to operate via its hydrolysis into fatty acids, which are then imported and resynthesized into TAG of ILI. For this, mycobacteria would require the presence of a TAG hydrolase at the surface of their cell wall. In addition, hydrolase activity would also be required inside mycobacteria when they degrade ILI as a source of energy. This implies either that two different lipases are involved or the dual location of a single lipase. Among the 24 putative lipases, LipY (Rv3097c) is the most likely candidate. It displays a PE (Pro-Glu) domain and plays a central role in TAG hydrolysis (43, 59). The PE domain contains a signal sequence for secretion of cytoplasmic LipY by the ESX-5 secretion system (53). After cleavage of its PE domain, LipY was seen to be inserted in the outermost layer of the cell wall (53, 54). At present, the possibility that other lipases, such as cutinase (61, 66) or monoglyceride lipases (67, 68), play a role in host TAG hydrolysis cannot be excluded.
Effects of lipid inclusions on mycobacterial replication and their ability to divide.The presence of nongrowing dormant bacilli is one of the hallmarks of asymptomatic TB. Dormancy implies a strongly reduced metabolism in addition to arrest of mycobacterial growth, but this has not been established for mycobacteria in FM of granulomas. Knowledge of the metabolic state into which mycobacteria enter once they accommodate themselves in FM has major relevance regarding mechanisms of dormancy and persistence. The present work found that M. avium in VLDL-driven FM accumulated ILI and stopped dividing. Mycobacterial growth was measured via CFU, which registered their number, but not their size. The latter was determined by EM. In spite of the division block, mycobacteria continued to grow at the same rate as in nonfoamy macrophages and became elongated as a result. The unchanged growth rate indicated that mycobacterial metabolism was not enhanced in the presence of the additional energy source in the form of ILI. This did not exclude the possibility of a qualitative change in metabolism in the form of an adaptation to the enhanced use of fat. It has indeed been shown that persistent mycobacteria upregulate genes involved in lipid metabolism (69, 70). Upon removal of VLDL, LB in host macrophages and ILI in mycobacteria disappeared. This allowed elongated mycobacteria to rapidly form division septa and divide until the same number of normal-sized mycobacteria had been established compared with their counterparts in lipid-free host cells. The present observations thus showed that the division block came into effect in the presence of ILI and was released upon their disappearance. More specifically, division was blocked only in the presence of ILI+3. Resumption of mycobacterial division occurred in correlation with the extent to which the relative abundance of ILI+3 profiles declined when VLDL, as the ultimate source of TAG for ILI, was removed (Fig. 9E).
Fluctuations of mycobacterial profiles between type ILI+3 and type ILI+2 are indicative of cell cycle events during replication.The kinetic observations of the formation and consumption of ILI+2 and ILI+3 suggested that the two types of ILI profiles coexisted in a precursor-product relationship (Fig. 2E and F and Fig. 7C, respectively). An individual mycobacterium could progress from type ILI+2 to ILI+3 during a net increase of its lipid content or could revert from type ILI+3 to ILI+2 during net lipid consumption. Its EM profile was either type ILI+2 or type ILI+3, but never a mixture of both. In the presence of sufficient TAG in the host cell, the relative abundance of ILI+3 profiles reached a steady state of about 45%, with ILI+2 profiles remaining abundant at about 30% (Fig. 2E). Assuming that this was not due to two different populations of mycobacteria with distinct characteristics for ILI formation, it suggested that equilibrium was established between the rates of ILI+3 formation and reversal to ILI+2.
The observation that only about 45% of mycobacteria were in the ILI+3 state at any one time and the indication that division was blocked only when mycobacteria were in this state raised the question of why the remaining 55% of mycobacteria were not found to divide. With about 20% of mycobacteria in the ILI+1/− state, their continued ability for normal growth and division would contribute about 20% to the increase in mycobacterial numbers. This could indeed have been the case as suggested by the slight increases in CFU counts that were observed toward the end of the 5-day-long exposures to VLDL (see the dotted line in Fig. 6A and the bottom graph in Fig. 8A). It remains to be determined whether this fraction of the bacterial population represents a different metabolic state and whether they do indeed divide normally or whether they are constantly derived from the remaining population. A similar argument does not hold for mycobacteria in the ILI+2 state, which make up about 35% of all bacteria at steady state. Their growth and division at the normal rate would have contributed an additional 35% to the rate of increase in CFU counts and should have become manifest after about 2 days of VLDL treatment but was not observed.
Why do mycobacteria in the ILI+2 state not seem able to divide? A mycobacterium might intermittently enter a phase of enhanced metabolism, during which protein synthesis and resulting elongation occurs (reminiscent of S phase in eukaryotes). This phase could involve enhanced consumption of TAG from ILI and their reversion from the ILI+3 state to the ILI+2 state. Elongation, however, would not be a time for septum formation and division. Before the latter sets in, TAG consumption would again slow down, allowing the reversal from the ILI+2 state to the ILI+3 state where septum formation and division would again be blocked. Support for this explanation came from the observation that division septa were never seen for ILI+2 profiles while VLDL, and thus sufficient TAG, was present. Only after removal of VLDL, during the decline of ILI+3, did ILI+2 profiles show division septa (Fig. 8C) because they could not revert to ILI+3 without sufficient TAG in the host cell. Evidence for cell cycle-specific phases in slow-growing bacteria has been reviewed, and a period of elongation has been distinguished between initiation and termination and division (Fig. 1 in reference 71).
Concluding remarks on the biological relevance of our model and future prospects.In general, the question must be asked to what extent model systems of FM, and in particular the present model of VLDL-driven FM, simulate conditions in foamy macrophages in granulomas. Insofar as the present observations do indeed resemble events in granulomas, it is more likely that they present a transitional phase toward a final stage of reduced metabolism and mycobacterial growth. Observations to this effect would require model systems that can simulate not only the biological environment in a granuloma but also a comparable length of time that bacilli would spend under these conditions.
Particular pathways for lipid metabolism may present potential targets for interference and drug development. In this context, we intend to use our experimental model in future work to investigate factors that influence macrophages as well as mycobacteria in their assimilation and disposal of lipid. This may provide the means to test various strategies aimed at sustaining dormancy or eradicating mycobacteria.
ACKNOWLEDGMENTS
This work was supported by core grants from the Institut National de la Santé et de la Recherche Médicale (INSERM) and the Centre National de la Recherche Scientifique (CNRS) to the Centre d'Immunologie de Marseille-Luminy (CIML), by grant ANR-09-MIEN-009-03 from the Agence Nationale de la Recherche (French National Research Agency) to I.C.-B., C.D.C., and S.C., and by funding from the Medical Research Council of South Africa to D.M.
We thank Jean Pierre Gorvel (CIML, CNRS UMR7280, INSERM UMR1104, Aix-Marseille University UM2, Marseille, France) and Laurent Kremer (CNRS UMR 5235, University of Montpellier, Montpellier, France) for continuous support and advice, Christophe Guilhot (IPBS, Toulouse, France) for providing the GFP-expressing M. avium strain, and the members of the PICsL confocal and electron microscopy core facilities (CIML-IBDML, Marseille, France) for expert technical assistance.
FOOTNOTES
- Received 24 September 2013.
- Returned for modification 12 October 2013.
- Accepted 14 November 2013.
- Accepted manuscript posted online 25 November 2013.
- Copyright © 2014, American Society for Microbiology. All Rights Reserved.
