Mycobacteria release active membrane vesicles that modulate immune responses in a TLR2-dependent manner in mice

Bacteria naturally release membrane vesicles (MVs) under a variety of growth environments. Their production is associated with virulence due to their capacity to concentrate toxins and immunomodulatory molecules. In this report, we show that the 2 medically important species of mycobacteria, Mycobacterium tuberculosis and Mycobacterium bovis bacille Calmette-Guérin, release MVs when growing in both liquid culture and within murine phagocytic cells in vitro and in vivo. We documented MV production in a variety of virulent and nonvirulent mycobacterial species, indicating that release of MVs is a property conserved among mycobacterial species. Extensive proteomic analysis revealed that only MVs from the virulent strains contained TLR2 lipoprotein agonists. The interaction of MVs with macrophages isolated from mice stimulated the release of cytokines and chemokines in a TLR2-dependent fashion, and infusion of MVs into mouse lungs elicited a florid inflammatory response in WT but not TLR2-deficient mice. When MVs were administered to mice before M. tuberculosis pulmonary infection, an accelerated local inflammatory response with increased bacterial replication was seen in the lungs and spleens. Our results provide strong evidence that actively released mycobacterial vesicles are a delivery mechanism for immunologically active molecules that contribute to mycobacterial virulence. These findings may open up new horizons for understanding the pathogenesis of tuberculosis and developing vaccines.

Many intracellular bacterial pathogens use membrane vesicles (MVs) as an alternative way to deliver ligands that can be recognized by host cells (1, 2). These MVs have been extensively studied in Gram-negative bacteria, which release MVs into the surrounding environment by a process that involves budding and “pinching off” from the outer membrane (3). The vesicles vary in size from 50 to 250 nm in diameter and contain phospholipids, proteins, and LPS as well as periplasmic components (3). Certain MVs can carry additional virulence factors, such as toxins, adhesins, or immunomodulatory compounds that are important for pathogenesis (1). Recently, a vesicular transport system has been described in the Gram-positive bacterium Bacillus anthracis (4) and in fungal pathogens whereby MVs function as so-called “virulence bags” to deliver molecules that modulate the host cell response (5). For mycobacteria, there is a report of MV production by Mycobacterium ulcerans, which implies that vesicular transport and delivery systems are also found among mycobacteria (6). Mycobacterium tuberculosis (Mtb) infects a large proportion of the human population, yet only a small percentage of those infected develop clinically apparent disease (7). This epidemiological fact implies that most human hosts mount effective responses to Mtb that control infection. The innate immune response against Mtb begins with the inhalation of bacilli into the lungs where it is recognized by the pattern recognition receptors (PRRs) of alveolar macrophages, dendritic cells, and monocytes. Moreover, the internalization of the bacteria leads to the activation and rapid production of proinflammatory cytokines and chemokines along with other molecules, such as MMPs, which are involved in the recruitment of neutrophils, natural killer and T cells, and granuloma formation (8, 9).

Recognition of Mtb by PRRs is mediated primarily by TLRs, especially TLR2, TLR4, and TLR9 (10). In fact, the host immune response to Mtb in macrophages has been reported to be mediated primarily by TLR2 (10). Bacilli in macrophage phagosomes secrete abundant cell-surface components that include a surprisingly large number of TLR2 ligands, such as bacterial lipoproteins (19-kDa lipoprotein, LprA, and LprG) and the immunologically active glycolipids LAM, PIM₂, and PIM₆ (11–15). Paradoxically, the TLR2 interactions with the Mtb ligands can activate functions that promote the killing of the bacteria but also appear to be part of the bacterium’s strategy to escape the host immune response. TLR2 lipid ligands traffic actively inside host macrophages and dendritic cells by incorporating into their endocytic networks and into exosomes (16). However, the mechanisms by which the bacilli shed these ligands are unknown.

In this study, we have demonstrated MVs in many mycobacterium species, including the medically important Mycobacterium bovis bacillus Calmette-Guérin (BCG) attenuated vaccine strain and virulent Mtb. We show that mycobacterial MVs transport lipids and proteins previously known to be involved in the subversion of the immune response of the host. In addition, we demonstrate that these vesicles can trigger an inflammatory response in a TLR2-dependent manner that directly modulates the outcome of the interaction with the host, contributing to virulence and pathogenesis in Mtb infection.

Release of MVs by multiple species of mycobacteria. We first investigated whether the M. bovis BCG and Mtb H37Rv strains release vesicles. We applied techniques developed for vesicle recovery from Gram-negative bacteria (17, 18) that used differential sedimentation and density gradient fractionation in an attempt to recover vesicles from mycobacterial culture supernatants. Transmission electron micrographic (TEM) analysis of pellets recovered from mycobacterial culture supernatants revealed vesicular structures associated to and pinching off the bacteria (Figure 1, A–D). The mass of this vesicular fraction was 250 μg ± 50 μg/l of minimal media (MM) (n = 4) and 135 μg ± 34 μg (n = 4) for BCG and Mtb, respectively. Electron micrographs revealed that more than 90% of the vesicles were closed forms (i.e., intact vesicles) (Figure 1, E and F), with diameters distributed between 60 and 300 nm (Figure 1, G and H).

Figure 1

Dependence of MV production by mycobacteria on cell viability. TEM showing production of MVs by BCG (A and C) or Mtb strain H37Rv (B and D). In A and B, vesicles are seen budding from the surface of the bacteria (arrows). In C and D, arrows indicate vesicles that appear to have separated from the bacterial cell. The inset shows a vesicle associated to the bacterial surface. Scale bars: 50 nm. (E and F) TEM of MVs isolated from a day-14 culture of BCG (E) or Mtb H37Rv (F) grown in MM. Scale bars: 100 nm. The size distribution of isolated vesicles from BCG (G) or Mtb H37Rv (H) was determined by electron microscopy. (I) BCG cultures were labeled with ¹⁴C-acetate (2 μCi ml^–1), harvested on day 5, and either left untreated or heat inactivated at 80°C for 2 hours. The viability of the cultures was determined by plating on 7H11 solid media at day 10, and percentage of survival was calculated by comparing CFU with bacterial particle counts obtained using a counting chamber. Each point represents an independent culture. (J) Cells prepared as described for I were resuspended in fresh medium and incubated for 5 additional days. The vesicle isolation procedure was carried out on supernatants from 3 independent cultures of heat-killed (80°C 2 hours) or live untreated BCG. The background level of radioactivity was determined isolating MVs on sterile medium containing ¹⁴C-acetate (2 μCi ml^–1).

Given that lipids released from dead or dying cells can self aggregate to form vesicles, we carefully assessed the viability of bacterial cultures from which vesicles were isolated and studied the effect of heat killing of the bacteria on vesicle generation. The viability of cells after 10 days of growth in 7H11 media was greater than 99%, as determined by comparing direct counts of bacterial particles with CFUs (Figure 1I). Although high viability argued against the vesicular material originating from dead cells, we could not completely exclude the possibility that vesicles originated from a minor fraction of dead or dying cells. Hence, we investigated alternate methods to establish an origin for the vesicles from live cells using metabolic labeling. Live and heat-inactivated mycobacterial cells previously radiolabeled with ¹⁴C-acetate (Merck) (19) were suspended in fresh MM and incubated for 5 additional days. The extent of the heat inactivation was determined by counting CFUs on 7H11 media (Figure 1I). Vesicles were isolated and radioactivity was determined as the measure of lipid synthesis. The level of radioactivity found in material recovered from heat-inactivated cultures was similar to the background recovered from sterile medium, whereas it was significantly greater in supernatants of cultures with live cells (Figure 1J). Furthermore, electron microscopy of mycobacteria revealed putative vesicles emerging from the bacterial cell wall with dimensions that were similar to isolated vesicles (Figure 1, A–D). We also investigated vesicle production in other mycobacterial species, including fast (Mycobacterium smegmatis, Mycobacterium phlei) and slow growers (Mycobacterium avium, Mycobacterium kansasii, and Mtb avirulent strain H37Ra) (Figure 2). All mycobacterial strains studied released vesicles in liquid cultures, and their size distribution as measured by TEM was similar, ranging from 40 to 250 nm in diameter (Figure 2).

Figure 2

Vesicle production is conserved in the Mycobacterium genus. Vesicles were isolated from cultures inoculated with the indicated strains and visualized by TEM. The size distribution of the vesicles was determined by TEM. The pictures are representative of 5 independent experiments. Scale bars: 100 nm.

Analysis of proteins associated with vesicles. To gain insight in the protein composition of MVs from both pathogenic (BCG and Mtb H37Rv) and nonpathogenic strains (M. smegmatis), vesicles were subjected to proteomic analysis using tryptic digestion prior to analysis by nano-LC-MS/MS. From 3 independent experiments, we identified 66 proteins in BCG, 48 in Mtb, and 64 in M. smegmatis vesicular preparations (Supplemental Tables 1–3; supplemental material available online with this article; doi: 10.1172/JCI44261DS1). All proteins detected were classified according to the functional category given by the databases TubercuList v2.3, BoviList v3.1, and SmegmaList v1.0 (Figure 3A). Overall, BCG and Mtb MVs had similar protein compositions, while M. smegmatis MVs contained a markedly different set of proteins. In Mtb, most vesicular proteins were classified as belonging to “cell wall and cell processes” (38%) or “intermediary metabolism and respiration” (19%). Proteins annotated as being involved in “lipid metabolism” were detected at a lower frequency (4%). Similarly, MVs BCG proteins belonging to cell wall and cell processes also represented 37%. Conversely, in M. smegmatis MVs, this category contained only 8% of the proteins identified. MVs from BCG were enriched in the “lipid metabolism” category (21% vs. 4% for Mtb and 5% for M. smegmatis; Figure 3A and Supplemental Tables 1–3). Remarkably, both BCG and Mtb MVs showed a prominent component of lipoproteins representing 20% and 10% of the total proteins, respectively (Tables 1 and 2). Since lipoproteins typically represent between 1% and 2% of mycobacterial genomes (20), this represented a major enrichment in the MVs. In order to experimentally confirm this enrichment in lipoproteins in MVs of BCG and Mtb, we analyzed the protein profiles of cellular lysates of these bacteria grown under the same culture conditions that were used for generation of MVs (Supplemental Tables 4 and 5). We found that lipoproteins represented 1.4% and 2.6% of the total cell-associated proteins in BCG and Mtb, respectively. This result established that MVs from BCG and Mtb strains were enriched in lipoproteins, a finding consistent with active release. In contrast, no lipoproteins were found in M. smegmatis MVs, despite the fact that M. smegmatis has 68 putative lipoproteins in the genome, representing 1.03% of the total protein in M. smegmatis (Supplemental Table 6).

Figure 3

Mycobacterium vesicle–associated proteins and lipids. (A) Pie charts showing the BCG, Mtb, and M. smegmatis vesicle–associated proteins grouped into functional categories as in TubercuList, BoviList, and SmegmaList. Note that the “lipid metabolism” functional category is increased in BCG vesicles and that “cell wall and cell processes”–associated proteins are overrepresented in both BCG and Mtb compared with M. smegmatis. (B) Vesicles from the supernatant of the BCG culture were loaded onto a SDS-polyacrylamide gel, blotted on a nitrocellulose membrane, and incubated with mAbs with the indicated specificities. (C) ImmunoGold electron microscopy of thin sections of BCG and Mtb H37Rv vesicles treated with the mAbs specific for the indicated proteins and detected with a 10-nm IgG gold-labeled anti-mouse antibody. Scale bars: 100 nm. (D) Total lipids of vesicles isolated from ¹⁴C-acetate–labeled cells were extracted with chloroform/methanol (2:1, v/v). Lipid extracts were separated by 2D TLC by applying an amount of lipids corresponding to 10,000 DPM per TLC plate and by using the solvent system E for separation of polar lipids. PG, phosphatidylglycerol; G, glycolipid; Ac₁PIM₂, monoacyl phosphatidylinositol dimannoside; Ac₂PIM₂, diacyl phosphatidylinositol dimannoside. Arrow indicates the origin.

The proteomic approach was validated by demonstrating the presence of some of the proteins identified by MS with serological techniques. In this regard, DnaK, 19 kDa (Lpqh), and LprG were detected by immunoblot analysis of vesicular preparations (Figure 3B) and visualized by electron microscopy as vesicle associated by ImmunoGold electron microscopy (Figure 3C). Lipoproteins 19 kDa and LprG were located at the vesicle membrane, as expected from their lipidic nature. In contrast, DnaK localized to the lumen of the vesicles. We detected the presence of known secreted compounds such as LAM or α-glucan in the vesicle preparations by ELISA (Supplemental Figure 1A). Based on TEM images, LAM was detected both associated with the vesicles and unassociated, while α-glucan was clearly outside and apparently not associated with the vesicles (Supplemental Figure 1B). Since the molecular weight of α-glucan is around 100 kDa, it can be retained in the membrane used to concentrate the supernatants. Consequently, we analyzed the fractions recovered after Optiprep density gradient purification by Western blot for the presence of LAM and by ELISA for α-glucan (Supplemental Figure 1C). The 19-kDa lipoprotein was used as a vesicle marker, and the presence of this protein corresponded to fractions containing LAM. In contrast, the α-glucan was associated with the heaviest fractions of the gradient consisting of protein aggregates and polysaccharides. In summary, MVs from pathogenic mycobacteria were enriched in lipoproteins, some of them well-known TLR2 ligands, along with other major cell surface compounds such as LAM, but did not contain α-glucan.

Analysis of vesicle-associated lipids. To evaluate the lipid profile of the vesicles, we performed 2D-TLC analyses of total lipid extracts from BCG and H37Rv (data not shown) vesicles isolated from ¹⁴C-acetate–labeled cultures. The vesicular lipids predominantly consisted of polar lipids (Figure 3D) with only low amounts of apolar lipids detectable (Supplemental Figure 2), consistent with the cell membrane being the likely origin of vesicles. The most abundant polar lipid species were phosphatidylinositol (PI) and acylated phosphatidylinositol dimannosides (PIM₂) along with cardiolipin (CL) and phosphatidylethanolamine (PE) (Figure 3D). To validate the 2D-TLC data, the BCG and Mtb vesicle samples were subjected to MALDI-TOF–MS analyses (Supplemental Figure 3). In addition to PI and acylated PIM₂, PIM₆ was also detected in the vesicle preparation. The results also showed that vesicles were enriched in polar lipids consistent with their origin from the cell membrane.

Production of MVs by BCG and Mtb during intracellular infection of macrophages in vitro and in vivo. The abundance of 19-kDa lipoprotein and LAM in vesicles provided an opportunity to use monoclonal antibodies to these components in immunoelectron microscopy for identification of vesicles in mycobacteria-infected macrophages in vitro and in vivo. This revealed vesicles shedding from the bacteria within phagosomes after 3 hours of infection (Figure 4A). In addition, MVs were seen trafficking at different locations in BCG-infected macrophages at 24 hours (Figure 4C). Similar findings were obtained with Mtb-infected macrophages at the same time intervals (Figure 4, B and D). To investigate whether the vesicle-release phenomenon could be documented in vivo, we infected mice with a high dose of CFSE-labeled bacteria. After 2 days, lungs were extracted and infected lung cells were sorted by FACS. Samples were processed for microscopy in the same manner as in vitro–infected macrophages and labeled with ImmunoGold. Images from this direct ex vivo analysis showed that BCG produced MVs within the phagosomes of macrophages that had been infected in vivo (Figure 4, E and F).

Figure 4

Production of vesicles by mycobacteria during intracellular infection of macrophages in vitro and in vivo. TEM showing MVs inside the BCG (A and C) or Mtb (B and D) in in vitro–infected macrophages at different time points after infection. Vesicles were observed by ImmunoGold labeling of sections with mAbs specific for 19-kDa lipoprotein (IT-12; 10 nm gold particles) and LAM (CS-35; 6 nm gold particles). (A and B) 19 kDa and LAM-labeled vesicles shedding from an intraphagosomal BCG (A) or H37Rv (B) after 3 hours of infection. (C) 19-kDa–labeled vesicle at the plasma membrane of a macrophage cell after 48 hours of in vitro infection with BCG. (D) TEM of a BMM infected in vitro with Mtb showing Mtb vesicles immunolabeled with IT-12 (10 nm) and CS-35 (6 nm) after 48 hours of infection. Scales bars: 50 nm. Pictures are representative of 2 independent experiments. (E and F) TEM of BCG in lung macrophages from infected mice revealing structures consistent with MV. As above, vesicles were observed by ImmunoGold labeling of sections with mAbs specific for 19 kDa (IT-12; 6 nm gold particles) and LAM (CS-35; 10 nm gold particles). Scale bars: 100 nm. Arrows indicate vesicles pinching off from the bacterial cell.

Vesicles stimulated macrophages in a TLR2-dependent manner. To evaluate the capacity of mycobacterial MVs to trigger a proinflammatory response, we added BCG vesicle preparations to freshly isolated BMM from WT and TLR2-knockout mice and measured cytokine and chemo­kine expression at different times after vesicle addition (Figure 5A). The TLR1/2 ligand Pam₃CSK₄ was used as a positive control. The MVs induced the production of IL-1β, IL-6, IL-10, IL-12, TNF, CXCL1, and MIP-1α/CCL3, and the majority of these cytokines were induced in a TLR2-dependent manner, since their production from MV-stimulated Tlr2^–/– cells was markedly reduced. However, induction of MIP-1α/CCL3 and IL-10 by MVs was preserved in Tlr2^–/– cells, indicating that MVs also contained components that could activate macrophages through receptors different from TLR2. In contrast, Pam₃CSK₄ induced all cytokines and chemokines studied in a TLR2-dependent manner. We were able to detect cytokines as early as 3 hours after MV stimulation, and the response peaked at 24 hours. The induction of MIP-1α/CCL3 and IL-10 by MVs had faster kinetics than the other inflammatory mediators, consistent with induction through a different, non–TLR2-related mechanism. Levels of IL-12 and TNF were not completely abolished in the TLR2-deficient macrophages. The residual levels of TNF could have been due to an alternative pathway of activation involving MIP-1α (21). The fact that the majority of cytokines remained persistently elevated at 48 hours suggested a prolonged effect of the MV-associated ligands. Subsequently, we analyzed the levels of cytokines produced by macrophages stimulated with the same amount of M. smegmatis MVs at the same time points and found detectable levels of cytokines including IL-10, IL-12, and IL-1β, but their concentrations were significantly lower than observed with BCG MVs. In addition, no chemokines or TNF were detected for M. smegmatis MV-stimulated macrophages at any time point analyzed in the study (Supplemental Figure 6A).

Figure 5

Responses of macrophages to BCG MVs. Freshly isolated BMM and alveolar macrophage cell line (AMJ2-C8) were stimulated with MVs and with Pam₃CSK₄ in vitro (500,000 cells/500 μl of media). Supernatants and cell lysates were collected at time points ranging from 0 to 48 hours after stimulation (x axes of all graphs indicate time in hours). (A) The supernatants of cultures of WT (white bars) or TLR2-knockout (black bars) isolated BMMs were assayed for cytokines IL-1β, IL-10, IL-12p70, TNF-α, and chemo­kines CXCL1 by multiplex ELISA (MSD Systems). Cytokines IL-6 and MIP-1α were analyzed by standard ELISA. Data represent mean ± SEM. (B) The levels of Cox-2 and MMP-9 were analyzed by Western blot and zymography, respectively. For Cox-2 analysis (top), lysates were made from cells (AMJ2-C8 alveolar macrophage cell line (AM), WT BMM, or Tlr2^–/– BMM collected at the indicated times, ranging from 0 to 48 hours after addition of PBS (unstimulated), Pam₃CSK₄, or BCG MVs. For measurement of MMP-9 activity, culture supernatants were collected at the indicated times following stimulation as shown. The graphs at the bottom show levels of MMP-9 activity relative to a recombinant MMP-9 protein standard. The data of the Cox-2 and MMP-9 analysis are representative of 3 independent experiments. Data represent mean ± SEM.

The cyclooxygenase Cox-2 is known to regulate production of MMP-9 in a TLR2-dependent manner (22). Since vesicles were enriched in TLR2 ligands, we analyzed the levels of Cox-2 in WT and TLR2-deficient BMM and alveolar macrophages by Western blot (Figure 5B). Again, Pam₃CSK₄ was used as a positive control for TLR2 stimulation. BCG MVs triggered expression of Cox-2 in a time-dependent manner in WT BMM, whereas Cox-2 was not induced in macrophages from Tlr2^–/– mice. The timing was very similar to that of induction of cytokines by MVs, with detectable levels appearing as early as 3 hours. The effect of MVs on Cox-2 expression was still present at 48 hours, indicating persistent stimulation by MVs as seen for cytokine secretion. Accordingly, the enhanced expression of Cox-2 protein resulted in increased levels of MMP-9 in the culture supernatants from BMMs stimulated with Pam₃CSK₄ or MVs, as measured by gelatin zymography (Figure 5B). The identity of the band was confirmed when the same experiment was carried out with BMMs from an mmp-9–knockout mouse, for which this gelatinolytic band was absent at 24 and 48 hours (Supplemental Figure 4). Neither Cox-2 nor MMP-9 was induced after stimulation of macrophages with M. smegmatis MVs (Supplemental Figure 6B). These results established that MVs from pathogenic mycobacteria were able to elicit production of a broad range of cytokines and other inflammatory mediators by macrophages in a TLR2-dependent manner.

Stimulation of TLR2-dependent inflammatory responses in the lung by MVs. To evaluate the effect of MVs on tissue inflammatory responses and the role for TLR2 in such responses, we administered a single intratracheal (i.t.) injection of MVs to WT or TLR2-deficient mice and evaluated the lungs at different times for inflammatory responses. On the fifteenth day after MV infusion, the lungs of the WT mice showed marked alterations compared with PBS-infused controls, including dense areas of mononuclear cell infiltration and interstitial pneumonia (Figure 6). In addition, the MV-injected mice also showed a predominantly diffuse alveolar damage pattern of injury with more compaction of alveolar spaces. In contrast, no pathological changes were detected in TLR2-deficient mice. By day 23, the lesions in lungs of WT mice had worsened, but there was still little inflammatory change in the lungs of MV-injected TLR2-knockout mice, which appeared similar to the lungs of PBS-treated controls. Lung inflammation still occurred when lower doses of vesicles were administered (Supplemental Figure 7). Administration of the same amount of M. smegmatis MVs also triggered an inflammatory response at day 15 and day 23 (Supplemental Figure 6C), but it was not as intense as observed with pathogenic MVs (Figure 6).

Figure 6

TLR2-dependent inflammation in mouse lungs after vesicle injection. Representative images of H&E-stained lung sections of i.t. vesicle or PBS-treated WT and Tlr2^–/– C57BL/6 mice. Lungs were analyzed at 15 and 23 days after treatment. The images are representative of 5 lungs. All pictures were taken using a Coolscope (Nikon). The experiment was done twice. Original magnification, ×40.

Since myeloid cells are known to be recruited as mediators of inflammatory processes in the lungs, we next asked whether i.t. injection of MVs was associated with changes in the numbers and phenotypic subsets of these myeloid cells. To analyze this, total lung cell suspensions were analyzed for the expression of CD11b, CD11c, and Gr1 (Figure 7A). A significant increase in the absolute numbers of pulmonary CD11b⁺Gr1^hi, CD11b⁺Gr1^int, and CD11b⁺CD11c^– cells (Figure 7B) was observed in the MV-treated relative to the PBS-treated WT mice. These increases were not observed in the TLR2-deficient mice that received i.t. MVs. The increase in the total number of cells expressing the CD11b marker was also apparent by confocal microscopy at day 15 (Figure 7C). When M. smegmatis MVs were administrated, we observed a significant increase in the absolute numbers of pulmonary CD11b⁺Gr1^hi population only on day 15, no change in CD11b⁺Gr1^int or CD11b⁺CD11c^– cells, and decrease in CD11b^–CD11c⁺ cells (Supplemental Figure 6D). These observations indicated that the vesicle treatment resulted in a profound alteration in the composition of the pulmonary myeloid populations in a TLR2-dependent manner. However, at day 23, only the CD11b⁺Gr1^hi cell population showed statistically significant differences (Supplemental Figure 5). Together, the above observations indicated that introduction of MVs into the lungs resulted in a profound alteration in the composition of the pulmonary myeloid populations in a TLR2-dependent manner.

Figure 7

TLR2-dependent increase in myeloid cells in the lungs following i.t. MV injection. (A) Representative flow cytometry dot plots showing the gating for analysis of different myeloid populations (R3, CD11b⁺Gr1^hi; R4, CD11b⁺Gr1^int; R5, CD11b^–CD11c⁺; and R6, CD11b⁺CD11c^–). (B) Bar graphs indicating the mean absolute numbers of the different populations of lung cells of WT (white bars) and Tlr2^–/– (black bars) mice. Data represent mean ± SEM. *P < 0.05; **P < 0.01; ***P < 0.001. (C) Confocal inmmunofluorescence microscopy of cryosections stained with DAPI and CD11b–Alexa Fluor 647 showing an increase in CD11b⁺ cells after vesicle treatment at 15 days in WT but not in the Tlr2^–/– mice. Data are representative of 2 independent experiments.

Vesicles impaired the control of Mtb infection. Next, we investigated whether exposure to MVs modified the course of experimental Mtb infection. Mice received an i.t. injection of either MVs or PBS and 15 days later were infected via aerosol with a low dose of Mtb H37Rv. Lung histology was examined at 15 and 23 days after infection (Figure 8A). As expected, inflammatory infiltrates were observed in lungs of both groups of animals at day 15, but the MV-treated mice showed more discrete granulomatous structures, suggesting a more rapid progression of the inflammatory response. The difference between MV-treated and untreated mice was more pronounced at day 23, with MV-treated mice showing larger and more numerous granulomas compared with the PBS-treated control mice (Figure 8C). In contrast, the amount of inflammation observed for mice given M. smegmatis MVs at days 15 and 23 was significantly lower compared with that of the BCG-treated mice (Supplemental Figure 6E).

Figure 8

Exacerbation of mycobacterial infection by pretreatment with isolated vesicles. (A) Representative images of H&E-stained lung sections of mice that received i.t. infusion of purified BCG MVs or PBS, followed by aerosol challenge with Mtb H37Rv 15 days later. Lungs were analyzed at 15 and 30 days after challenge. (B) Representative images of H&E-stained lung sections of BCG-vaccinated WT C57BL/6 mice infused with vesicles or PBS and challenged with a low dose of H37Rv (50–100 CFUs). Mice were vaccinated with BCG by intradermal injection and 6 weeks later received i.t. PBS (control) or purified BCG MVs. Aerosol challenge with MTb H37Rv was done 15 days later. Lungs were analyzed at 30 days after the challenge. All pictures were taken using a digital camera Coolscope (Nikon). Original magnification, ×40. (C) Quantification of the number of granuloma per lung in mice receiving the indicated treatments followed by Mtb H37Rv challenge. Analysis was done at day 30 after challenge. (D and E) CFUs in lungs (D) and spleens (E) of mice receiving the indicated treatments followed by Mtb H37Rv challenge. Analysis was done at day 30 after challenge. Experimental data are representative of 2 independent experiments. Data represent mean ± SEM. **P < 0.01.

We also studied BCG-vaccinated mice to determine whether vesicles can alter the outcome of the infection in animals with a recall response induced by previous vaccination (Figure 8B). The BCG-vaccinated mice that received MV injections showed markedly increased granulomatous inflammation (Figure 8B), and the number of granulomas per lung was higher compared with BCG-vaccinated mice that received only PBS injections (Figure 8C). The number of CFUs in the lung of unvaccinated mice at day 23 following Mtb infection was significantly higher in mice treated with vesicles (Figure 8D). This was also true when CFUs were counted in spleen (Figure 8E), suggesting that vesicles can promote the dissemination of the bacilli form the lung to other lymphoid organs. Similar trends were observed for BCG-vaccinated mice, although this did not reach statistical significance (Figure 8D). Conversely, the bacterial counts in the lung of mice treated with M. smegmatis MVs were significantly lower at day 23 compared with the PBS-treated mice, suggesting an alternative response that may be beneficial for the host (Supplemental Figure 6F) and indicating specificity in the inflammatory response triggered by pathogenic mycobacterial MVs.

The present work represents the first study, to our knowledge, of production of MVs in 2 of the most clinically significant species of mycobacteria, M. bovis BCG and Mtb, along with other representatives of the mycobacterial genus including nonpathogenic strains. The results extend a prior report on vesicle production by M. ulcerans (6) and further define a newly recognized aspect of the host-pathogen interaction in mycobacterial infections with clear implications for pathogenesis, since MVs from nonpathogenic strains do not drive the same responses in the host. Furthermore, our results complement the observation that another Gram-positive bacterium, B. anthracis, produces MV (4), extending this mechanism into the subset of the Gram-positive group defined by the mycobacteria. Our results indicated that BCG and Mtb vesicles were physiologic products of live bacilli growing in liquid culture or within macrophages. Using different independent lines of experimentation, we showed that the production of vesicles by BCG depended on cell viability. Thus, vesicles were recovered only from live cultures, vesicle fractions were metabolically labeled only by live cells, and vesicular structures could be identified on the surface of mycobacteria in electron micrographs. Hence, we conclude that mycobacterial species, such as Gram-negative bacteria (1) and fungi (5), produce MVs that may facilitate the export of a variety of potential virulence factors.

The earlier study of vesicles released by M. ulcerans showed these to be enriched in mycolactone, a known toxin produced by this microorganism (6). This suggests that like many Gram-negative bacterial pathogens, mycobacteria produce MVs as carriers for virulence factors that modify the responses of host cells to infection (1). Consistent with this idea, our proteomic analyses of BCG, Mtb, and M. smegmatis MVs showed that only BCG and Mtb MVs were enriched in proteins associated with virulence, including a remarkably high content of putative TLR2 ligands such as the lipoproteins 19 kDa, LprA, and LprG. These molecules have been shown to downregulate the levels of MHCII molecules on macrophages and dendritic cells through a mechanism involving recognition by TLR2 (11–13). The lipid profile of BCG and Mtb MVs suggested that they originated from the bacterial cell membrane, since polar lipids typically found in the plasma membrane were the most abundant species. Interestingly, some of these polar lipids, such as PIM₂ or PIM₆, are known to be ligands of TLR2. As suggested by the nature of their protein and lipid cargo, mycobacterial MVs produced a profound TLR2-dependent inflammatory response in vitro and in vivo. In addition, MVs induced a range of inflammatory cytokines as well as the expression of key molecules implicated in early events of granuloma formation such as Cox-2 or MMP-9. Our in vivo studies indicated that MVs stimulated myeloid cell recruitment, leading to an increase in the CD11b⁺Gr1^int population subset in the lung. Remarkably, when vesicle-treated mice were challenged with Mtb H37Rv, they showed an increase in granulomatous inflammation in the lungs as well as an increase in the number of bacilli in both their lungs and spleens.

Mycobacterial lipids, including those originating from the cell wall, and proteins shed by phagocytosed mycobacteria are incorporated into released vesicles known as exosomes, which are produced by exocytic machinery of the mammalian host cells (16, 23). The way in which these surface compounds are released from the bacteria remains elusive. Given our observations using TEM showing the production of vesicles by the bacteria within the phagosomes of BMMs in vitro and in vivo and BMDCs (Rafael Prados-Rosales, unpublished observations), we propose that direct release by the bacteria of MVs provides a mechanism for transport of immunomodulatory compounds such as lipoproteins. This may lead to the distribution of these compounds to various compartments within infected host cells or to their incorporation into host cell–derived exosomes. Thus, generation and release of MVs may allow mycobacteria confined within phagosomes to deliver virulence factors to other cellular compartments or into the extracellular milieu to affect neighboring cells.

In addition to lipoproteins, our lipidomic studies showed that PI and PIM₂ along with CL and PE were the most prevalent polar lipids in MVs. Moreover, LAM was also detected as a part of the vesicles. LAM, lipomannan (LM), and PIMs are major lipoglycans that are noncovalently attached to the plasma membrane through their phosphatidyl-myo-inositol anchor and extend to the exterior of the cell wall (24). LAM is an important ligand in the interaction between Mtb and macrophages (25) and contributes to modulating the inflammatory response by downregulating IL-12 production (26). In addition, LAM inhibits phosphadityl inositol-3-phosphate (PI3P) association with Mtb phagosomes (27) and blocks calcium flux in infected cells (28). Conversely, LM and PIMs can induce a proinflammatory response by inducing TNF-α, IL-8, and IL-12 secretion (29). Other important mycobacterial cell wall components, such as the glycolipid trehalose dimycolate (TDM), have been implicated in the development of an inflammatory response and in granuloma formation (30, 31). However, we did not find TDM associated with MVs in our 2D-TLC analysis. Recently, mycobacterial PIM₂ was reported to induce the expression of Cox-2 and subsequently MMP-9 in a TLR2-dependent manner (22), which parallels our findings in cells exposed to isolated MVs. These enzymes modify the host response, as Cox-2 generates prostaglandins that modulate the cell death response to infection (32) and MMP-9 is a metalloprotease known to be involved in the remodeling of the extracellular matrix during the early events of granuloma formation (33, 34). We did not explore the contribution of other TLRs, such as TLR4, TLR6, or TLR9, which have been shown to have a role in the recognition of Mtb in mice and human (35–38). However, we detected DNA in purified and intact mycobacterial MVs (our unpublished data), suggesting a possible interaction of MVs with TLR9.

One of the key features of mycobacterial infections is the production of an inflammatory Th1 response that promotes the secretion of selected cytokines such as TNF-α, INF-γ, and IL-12 (39, 40). However, our analysis of cytokines produced by BMMs stimulated with MVs did not indicate the induction of a typical Th1 priming milieu, since cytokines that oppose Th1 priming such as IL-10 were also induced (41). IL-10 has been shown to block activation of macrophages, which may increase the dissemination of the bacteria. In addition, the chemokines CXCL1 and CCL3 were induced in BMMs by MVs, which along with other cytokines are critically involved in recruitment of multiple effector cells at the site of infection. The analysis of the type of myeloid cells recruited after a BCG MV treatment showed an increase in the CD11b⁺Gr1^hi, CD11b⁺ Gr1^int, and CD11b⁺CD11c^– cell populations. This is relevant because a strong inflammatory response and an increase in the cell population CD11b⁺Gr1^int have been recently associated with progression of tuberculosis (42), providing support for the hypothesis that MVs contribute to the virulence of Mycobacterium. Moreover, this population has been shown to be permissive for mycobacterial replication in mice (43). Although more phenotypic markers are required to define this CD11b⁺Gr1^int population, we believe that they may be a subset of myeloid-derived suppressor cells (MDSC), which may have inhibitory properties in the context of infection by intracellular bacterial pathogens, including Mtb (44). Remarkably, this CD11b⁺Gr1^int myeloid cell population did not increase upon treatment with M. smegmatis MVs, suggesting that its recruitment may be specifically associated with pathogenic mycobacteria. We note that there is increasing evidence that macrophage and dendritic cell responses to microbial antigens can differ (45) and the interaction of vesicles with dendritic cells should be a priority for future studies.

Our findings point to MVs as a factor that is detrimental for the host during Mtb infection. In support of this, we note the observations that mice pretreated with pathogenic MVs developed a more acute inflammation and higher bacterial loads in the lungs compared with control and nonpathogenic MV-treated mice and that preexposure to MVs also increased dissemination of bacilli from the lung to other lymphoid organs such as the spleen. Furthermore, we observed MVs in mycobacteria recovered from mice, providing evidence for the formation of these structures in vivo and lending support to the view that MVs represent a bacterial release mechanism relevant to pathogenesis.

In summary, we have confirmed the existence of a vesicular system in mycobacteria and demonstrate that the vesicles produced contain numerous components associated with virulence. The finding that MVs are released inside macrophages in vivo suggests that they have the capacity to influence the outcome of the macrophage-mycobacteria interaction. Mycobacterial vesicles are proinflammatory through a TLR2-dependent mechanism, providing additional evidence for the key role of the innate immune receptor in host defense against Mtb. The work reported in the current study raises many important questions about the relevance of MVs to the adaptive immune response against Mtb, which ultimately may be resolved through the isolation of bacterial mutants that are deficient in vesicle production. The discovery of a vesicular transport system in mycobacteria opens up new horizons for the pathogenesis of tuberculosis and vaccine development.

View Supplemental data

A. Casadevall is supported by NIH awards HL059842, AI033774, AI052733, AI033142, and the AERAS foundation for vaccine development. A. Casadevall is also supported by the Northeastern Biodefense Center under grant U54-AI057158-Lipkin. S.A. Porcelli and W.R. Jacobs Jr. were supported by the NIH Program Project grant AI-063537. R. Kalscheuer acknowledges support from the Juergen Manchot Foundation. Flow cytometry studies were carried out using core facilities supported by the Einstein Cancer Center (NIH/NCI CA013330) and the Einstein Center for AIDS Research (NIH AI-51519). Special thanks are given to C.A. Duque and V. Cabezon for their helpful discussion, support, and inspiration.

Address correspondence to: Arturo Casadevall, Albert Einstein College of Medicine, Jack and Pearl Resnick Campus, 1300 Morris Park Avenue, Forchheimer Building, Room 411, New York, New York 10461, USA. Phone: 718.430.2811; Fax: 718.430.8771; E-mail: casadeva@aecom.yu.edu.

Conflict of interest: The authors have declared that no conflict of interest exists.

Reference information: J Clin Invest. 2011;121(4):1471–1483. doi:10.1172/JCI44261.

Luis Martinez’s present address is: Long Island University-C.W. Post, Department of Biomedical Sciences, Brookville, New York, USA.

1. Kuehn MJ, Kesty NC. Bacterial outer membrane vesicles and the host-pathogen interaction. Genes Dev. 2005;19(22):2645–2655.
   View this article via: PubMed CrossRef Google Scholar
2. Deatherage BL, Lara JC, Bergsbaken T, Rassoulian Barrett SL, Lara S, Cookson BT. Biogenesis of bacterial membrane vesicles. Mol Microbiol. 2009;72(6):1395–1407.
   View this article via: PubMed Google Scholar
3. Beveridge TJ. Structures of gram-negative cell walls and their derived membrane vesicles. J Bacteriol. 1999;181(16):4725–4733.
   View this article via: PubMed Google Scholar
4. Rivera J, Cordero RJ, Nakouzi AS, Frases S, Nicola A, Casadevall A. Bacillus anthracis produces membrane-derived vesicles containing biologically active toxins. Proc Natl Acad Sci U S A. 2010;107(44):19002–19007.
   View this article via: PubMed CrossRef Google Scholar
5. Rodrigues ML, et al. Extracellular vesicles produced by Cryptococcus neoformans contain protein components associated with virulence. Eukaryot Cell. 2008;7(1):58–67.
   View this article via: PubMed CrossRef Google Scholar
6. Marsollier L, et al. Impact of Mycobacterium ulcerans biofilm on transmissibility to ecological niches and Buruli ulcer pathogenesis. PLoS Pathog. 2007;3(5):e62.
   View this article via: PubMed CrossRef Google Scholar
7. Bloom BR, Fine PEM. The BCG experience: Implications for future vaccines against tuberculosis. In: Bloom BR, ed. Tuberculosis, Pathogenesis, Protection and Control . Washington, DC, USA: American Society for Microbiology Press; 1994:531–557.
8. Davis JM, Ramakrishnan L. The role of the granuloma in expansion and dissemination of early tuberculous infection. Cell. 2009;136(1):37–49.
   View this article via: PubMed CrossRef Google Scholar
9. Feldmesser M, Casadevall A. Effect of serum IgG1 to Cryptococcus neoformans glucuronoxylomannan on murine pulmonary infection. J Immunol. 1997;158(2):790–799.
   View this article via: PubMed Google Scholar
10. Jo EK, Yang CS, Choi CH, Harding CV. Intracellular signalling cascades regulating innate immune responses to Mycobacteria: branching out from Toll-like receptors. Cell Microbiol. 2007;9(5):1087–1098.
    View this article via: PubMed CrossRef Google Scholar
11. Brightbill HD, et al. Host defense mechanisms triggered by microbial lipoproteins through toll-like receptors. Science. 1999;285(5428):732–736.
    View this article via: PubMed CrossRef Google Scholar
12. Pecora ND, Gehring AJ, Canaday DH, Boom WH, Harding CV. Mycobacterium tuberculosis LprA is a lipoprotein agonist of TLR2 that regulates innate immunity and APC function. J Immunol. 2006;177(1):422–429.
    View this article via: PubMed Google Scholar
13. Gehring AJ, Dobos KM, Belisle JT, Harding CV, Boom WH. Mycobacterium tuberculosis LprG (Rv1411c): a novel TLR-2 ligand that inhibits human macrophage class II MHC antigen processing. J Immunol. 2004;173(4):2660–2668.
    View this article via: PubMed Google Scholar
14. Gilleron M, Ronet C, Mempel M, Monsarrat B, Gachelin G, Puzo G. Acylation state of the phosphatidylinositol mannosides from Mycobacterium bovis bacillus Calmette Guerin and ability to induce granuloma and recruit natural killer T cells. J Biol Chem. 2001;276(37):34896–34904.
    View this article via: PubMed CrossRef Google Scholar
15. Means TK, Wang S, Lien E, Yoshimura A, Golenbock DT, Fenton MJ. Human toll-like receptors mediate cellular activation by Mycobacterium tuberculosis. J Immunol. 1999;163(7):3920–3927.
    View this article via: PubMed Google Scholar
16. Beatty WL, Rhoades ER, Ullrich HJ, Chatterjee D, Heuser JE, Russell DG. Trafficking and release of mycobacterial lipids from infected macrophages. Traffic. 2000;1(3):235–247.
    View this article via: PubMed CrossRef Google Scholar
17. Bauman SJ, Kuehn MJ. Purification of outer membrane vesicles from Pseudomonas aeruginosa and their activation of an IL-8 response. Microbes Infect. 2006;8(9–10):2400–2408.
    View this article via: PubMed CrossRef Google Scholar
18. McBroom AJ, Kuehn MJ. Release of outer membrane vesicles by Gram-negative bacteria is a novel envelope stress response. Mol Microbiol. 2007;63(2):545–558.
    View this article via: PubMed CrossRef Google Scholar
19. Kremer L, et al. Identification and structural characterization of an unusual mycobacterial monomeromycolyl-diacylglycerol. Mol Microbiol. 2005;57(4):1113–1126.
    View this article via: PubMed Google Scholar
20. Rezwan M, Grau T, Tschumi A, Sander P. Lipoprotein synthesis in mycobacteria. Microbiology. 2007;153(pt 3):652–658.
    View this article via: PubMed CrossRef Google Scholar
21. Fahey TJ 3rd, et al. Macrophage inflammatory protein 1 modulates macrophage function. J Immunol. 1992;148(9):2764–2769.
    View this article via: PubMed Google Scholar
22. Bansal K, Kapoor N, Narayana Y, Puzo G, Gilleron M, Balaji KN. PIM2 Induced COX-2 and MMP-9 expression in macrophages requires PI3K and Notch1 signaling. PLoS One. 2009;4(3):e4911.
    View this article via: PubMed CrossRef Google Scholar
23. Giri PK, Schorey JS. Exosomes derived from M. Bovis BCG infected macrophages activate antigen-specific CD4+ and CD8+ T cells in vitro and in vivo. PLoS One. 2008;3(6):e2461.
    View this article via: PubMed CrossRef Google Scholar
24. Besra GS, Morehouse CB, Rittner CM, Waechter CJ, Brennan PJ. Biosynthesis of mycobacterial lipoarabinomannan. J Biol Chem. 1997;272(29):18460–18466.
    View this article via: PubMed CrossRef Google Scholar
25. Schlesinger LS, Hull SR, Kaufman TM. Binding of the terminal mannosyl units of lipoarabinomannan from a virulent strain of Mycobacterium tuberculosis to human macrophages. J Immunol. 1994;152(8):4070–4079.
    View this article via: PubMed Google Scholar
26. Nigou J, Zelle-Rieser C, Gilleron M, Thurnher M, Puzo G. Mannosylated lipoarabinomannans inhibit IL-12 production by human dendritic cells: evidence for a negative signal delivered through the mannose receptor. J Immunol. 2001;166(12):7477–7485.
    View this article via: PubMed Google Scholar
27. Fratti RA, Backer JM, Gruenberg J, Corvera S, Deretic V. Role of phosphatidylinositol 3-kinase and Rab5 effectors in phagosomal biogenesis and mycobacterial phagosome maturation arrest. J Cell Biol. 2001;154(3):631–644.
    View this article via: PubMed CrossRef Google Scholar
28. Vergne I, Chua J, Deretic V. Tuberculosis toxin blocking phagosome maturation inhibits a novel Ca2+/calmodulin-PI3K hVPS34 cascade. J Exp Med. 2003;198(4):653–659.
    View this article via: PubMed CrossRef Google Scholar
29. Dao DN, et al. Mycobacterium tuberculosis lipomannan induces apoptosis and interleukin-12 production in macrophages. Infect Immun. 2004;72(4):2067–2074.
    View this article via: PubMed Google Scholar
30. Guidry TV, Hunter RL Jr, Actor JK. Mycobacterial glycolipid trehalose 6,6’-dimycolate-induced hypersensitive granulomas: contribution of CD4+ lymphocytes. Microbiology. 2007;153(pt 10):3360–3369.
    View this article via: PubMed CrossRef Google Scholar
31. Rhoades ER, Geisel RE, Butcher BA, McDonough S, Russell DG. Cell wall lipids from Mycobacterium bovis BCG are inflammatory when inoculated within a gel matrix: characterization of a new model of the granulomatous response to mycobacterial components. Tuberculosis (Edinb). 2005;85(3):159–176.
    View this article via: PubMed Google Scholar
32. Chen M, et al. Lipid mediators in innate immunity against tuberculosis: opposing roles of PGE2 and LXA4 in the induction of macrophage death. J Exp Med. 2008;205(12):2791–2801.
    View this article via: PubMed CrossRef Google Scholar
33. Taylor JL, et al. Role for matrix metalloproteinase 9 in granuloma formation during pulmonary Mycobacterium tuberculosis infection. Infect Immun. 2006;74(11):6135–6144.
    View this article via: PubMed CrossRef Google Scholar
34. Volkman HE, Pozos TC, Zheng J, Davis JM, Rawls JF, Ramakrishnan L. Tuberculous granuloma induction via interaction of a bacterial secreted protein with host epithelium. Science. 2010;327(5964):466–469.
    View this article via: PubMed CrossRef Google Scholar
35. Abel B, et al. Toll-like receptor 4 expression is required to control chronic Mycobacterium tuberculosis infection in mice. J Immunol. 2002;169(6):3155–3162.
    View this article via: PubMed Google Scholar
36. Bafica A, Scanga CA, Feng CG, Leifer C, Cheever A, Sher A. TLR9 regulates Th1 responses and cooperates with TLR2 in mediating optimal resistance to Mycobacterium tuberculosis. J Exp Med. 2005;202(12):1715–1724.
    View this article via: PubMed CrossRef Google Scholar
37. Sugawara I, Yamada H, Li C, Mizuno S, Takeuchi O, Akira S. Mycobacterial infection in TLR2 and TLR6 knockout mice. Microbiol Immunol. 2003;47(5):327–336.
    View this article via: PubMed Google Scholar
38. Mayer-Barber KD, et al. Caspase-1 independent IL-1beta production is critical for host resistance to mycobacterium tuberculosis and does not require TLR signaling in vivo. J Immunol. 2010;184(7):3326–3330.
    View this article via: PubMed CrossRef Google Scholar
39. Chensue SW, Warmington K, Ruth J, Lincoln P, Kuo MC, Kunkel SL. Cytokine responses during mycobacterial and schistosomal antigen-induced pulmonary granuloma formation. Production of Th1 and Th2 cytokines and relative contribution of tumor necrosis factor. Am J Pathol. 1994;145(5):1105–1113.
    View this article via: PubMed Google Scholar
40. Chensue SW, Warmington KS, Ruth JH, Lincoln P, Kunkel SL. Cytokine function during mycobacterial and schistosomal antigen-induced pulmonary granuloma formation. Local and regional participation of IFN-gamma, IL-10, and TNF. J Immunol. 1995;154(11):5969–5976.
    View this article via: PubMed Google Scholar
41. Gordon S. Alternative activation of macrophages. Nat Rev Immunol. 2003;3(1):23–35.
    View this article via: PubMed CrossRef Google Scholar
42. Lyadova IV, et al. In mice, tuberculosis progression is associated with intensive inflammatory response and the accumulation of gr-1 cells in the lungs. PLoS One. 2010;5(5):e10469.
    View this article via: PubMed CrossRef Google Scholar
43. Antonelli LR, et al. Intranasal Poly-IC treatment exacerbates tuberculosis in mice through the pulmonary recruitment of a pathogen-permissive monocyte/macrophage population. J Clin Invest. 2010;120(5):1674–1682.
    View this article via: JCI PubMed CrossRef Google Scholar
44. Gabrilovich DI, Nagaraj S. Myeloid-derived suppressor cells as regulators of the immune system. Nat Rev Immunol. 2009;9(3):162–174.
    View this article via: PubMed Google Scholar
45. Harding CV, Boom WH. Regulation of antigen presentation by Mycobacterium tuberculosis: a role for Toll-like receptors. Nat Rev Microbiol. 2010;8(4):296–307.
    View this article via: PubMed CrossRef Google Scholar
46. Schwebach JR, et al. Glucan is a component of the Mycobacterium tuberculosis surface that is expressed in vitro and in vivo. Infect Immun. 2002;70(5):2566–2575.
    View this article via: PubMed CrossRef Google Scholar
47. Besra GS. Preparation of cell-wall fractions from mycobacteria. Methods Mol Biol. 1998;101:91–107.
    View this article via: PubMed Google Scholar
48. Laval F, Laneelle MA, Deon C, Monsarrat B, Daffe M. Accurate molecular mass determination of mycolic acids by MALDI-TOF mass spectrometry. Anal Chem. 2001;73(18):4537–4544.
    View this article via: PubMed CrossRef Google Scholar
49. Takeuchi O, et al. Differential roles of TLR2 and TLR4 in recognition of gram-negative and gram-positive bacterial cell wall components. Immunity. 1999;11(4):443–451.
    View this article via: PubMed CrossRef Google Scholar
50. Roberts AD, Belisle JT, Turner J, Gonzalez-Juarrero M, Cooper AM, Orme IM. Murine models of tuberculosis. In: Kaufmann SHE, Kabelitz D, eds. Methods in Microbiology . Vol. 32. Burlington, Massachusetts, USA: Academic Press Ltd; 2002:433–462 .