Exosomes Released from Infected Macrophages Contain Mycobacterium avium Glycopeptidolipids and Are Proinflammatory*

Mycobacterium avium is a major opportunistic pathogen in HIV-positive individuals and is responsible for increased morbidity and mortality in AIDS patients. M. avium express glycopeptidolipids (GPLs) as a major cell wall constituent, and recent studies suggest that GPLs play an important role in M. avium pathogenesis. In the present study we show that M. avium-infected macrophages release GPLs, which are trafficked from the phagosome through the endocytic network to multivesicular bodies. Prior studies have shown that multivesicular bodies can fuse with the plasma membrane releasing small 50 to 100 nm vesicles known as exosomes. We found that M. avium-infected macrophages release exosomes containing GPLs leading to the transfer of GPLs from infected to uninfected macrophages. Interestingly, exosomes isolated from M. avium-infected but not from uninfected macrophages can stimulate a proinflammatory response in resting macrophages. This proinflammatory response is dependent on Toll like receptor (TLR) 2, TLR4, and MyD88 suggesting that released exosomes contain M. avium-expressed TLR ligands. Our studies are the first to demonstrate that exosomes isolated from mycobacteria-infected macrophages can induce a proinflammatory response, and we hypothesize that exosomes play an important role in immune surveillance during intracellular bacteria infections.

GPL significantly attenuates the virulence of M. avium (7,8). However, it is not clear how GPL can modulate the immune response and therefore accentuate the survival of M. avium in the host.
The GPLs consist of a tripeptide amino alcohol core modified with an amide-linked fatty acid, a methylated rhamnose and a 6-deoxytalose (9). The M. avium GPLs can be further modified in the length and composition of sugars attached to the 6-deoxytalose residue. The GPLs are non-covalently attached to the cell wall of mycobacteria. Interestingly, studies by Vergne et al. (10) showed that GPLs can interact with host membranes and suggests that GPLs could promote mycobacteria survival by interfering with the membrane-mediated functions of the host cells. There is also evidence showing that GPLs accumulate inside infected cells (11)(12)(13). Moreover, due to the extensive movement of membrane within the endocytic pathway, the mycobacterial GPL could be trafficked to other cellular compartments and modulate cellular functions.
In the present study, we hypothesized that the mycobacterial GPL are released from the M. avium surface and trafficked throughout the cell as has been shown for mannosylated LAM and other mycobacterial lipids (14). However, our studies indicate that GPLs are more restrictive in their distribution; with the majority of the GPL trafficking to a distinct compartment within the endocytic pathway designated the multivesicular body or MVB. Prior studies have shown that fusion of MVBs with the plasma membrane results in the release of small 50 -100 nm vesicles called exosomes (15). We found that exosomes released from M. avium-infected macrophages contain GPLs and that these exosomes can interact with uninfected macrophage leading to retention of GPLs in these "bystander" cells. Interestingly, exosomes purified from the culture supernatant of M. avium but not from uninfected cells induced a proinflammatory response in exosome-treated macrophages. To our knowledge, our studies are first to show that the exosomes carrying bacterial components could induce a proinflammatory response and suggests a novel mechanism by which stimulator molecules present on intracellular pathogens can be released from infected cells to promote an immune response.
Bone marrow-derived macrophages (BMMs), used in all experiments, was isolated from 6-to 8-week-old BALB/c mice as previously described (16). The macrophages were used 7-14 days after isolation or frozen after 7 days of culture in freezing media (50% Dulbecco's modified Eagle's medium, 40% fetal bovine serum, and 10% endotoxin-tested Me 2 SO (Sigma). Thawed or fresh macrophages were cultured on non-tissue culture plates for 3-7 days and then re-plated at 3 ϫ 10 5 cells per 35-mm tissue culture plate. The cells were allowed to adhere for 24 h prior to treatments with mycobacteria or inhibitors. The plates were then incubated at 37°C in 5% CO 2 for various times. The infection was done for 4 h, and then the media on the cells was replaced with fresh media. All tissue culture reagents were found to be negative for the endotoxin contamination by the E-Toxate assay (Sigma).
Bacteria Culture-To generate the M. avium 2151 variant stocks, the bacteria were grown on Middlebrooks 7H11 plates for 20 -30 days. The different morphotypes of M. avium 2151 were then resuspended in freezing media (Middlebrooks 7H9, 10% glycerol, glucose, oleic acids, Tween 20, and NaCl (Ref. 16). All the stocks were quantitated by the serial dilutions. M. avium 101 and M. avium A5 were also prepared as described for M. avium 2151. Infection assays evaluated by fluorescence microscopy was performed on each stock of mycobacteria to determine the infection ratio required to obtain ϳ80% of the macrophages infected. The amount of mycobacteria required to obtain this infection rate varied between species and batch preps of mycobacteria. However, the results were consistent across the different mycobacteria-to-macrophage ratios when the infection levels were kept constant.
Complement Opsonization-Appropriate concentration of mycobacteria were suspended in macrophage culture media containing 10% horse serum as a source of complement components and incubated for 2 h at 37°C (16).
TLC Immunostaining-The TLC immunostaining of GPLs on TLC was performed according to the method of Watarai et al. (18) with slight modifications. Briefly, alkali-stable GPLs were separated by TLC using Silica Gel-60 plates (EM Science) with chloroform/methanol/water (30:8:1, v/v) as the developing solvent. The GPLs were visualized using a-naphthol/sulfuric acid as the spray reagent. The dried plate was soaked for 1 min in a 0.02% solution of polyisobutylmethacrylate (Glycotech, MD) dissolved in acetone, allowed to air dry, and then blocked by incubation in PBS containing 10% heat-inactivated horse serum at 37°C for 2 h. The plate was then rinsed with PBS containing 0.1% Tween 20 (washing buffer) and incubated with anti-GPL monoclonal antibody in PBS for 2 h at room temper-ature. Following this, the plate was washed thrice with washing buffer and probed with horseradish peroxidase-conjugated anti-mouse immunoglobulin G antiserum (Amersham Biosciences) at room temperature for 2 h. The plate was washed, and the bound Abs was detected using SuperSignal West Femto enhanced chemiluminescence reagents (Pierce).
Isolation of Exosomes and Apoptotic Vesicles-The fetal calf serum used in the cell culture media for exosome isolation was centrifuged at 100,000 ϫ g, 15 h to remove any contaminating exosomes from the media. J774 cells were infected with M. avium 2151 SmT for 4 h and washed extensively to remove extracellular bacteria. After 72 h, the culture medium was collected and centrifuged twice at 300 ϫ g, 10 min to remove whole cells, followed by centrifugation at 1200 ϫ g for 10 min to remove any bacilli. The supernatant from the previous spins was centrifuged at 10,000 ϫ g for 30 min. This was followed with ultracentrifugation at 100,000 ϫ g for 1 h. The resulting pellet following 100,000 ϫ g centrifugation was further purified by sucrose gradient. Apoptotic vesicles were purified by consecutive centrifugations as described previously (19,20). Briefly, apoptosis was induced by fetal calf serum deprivation (21,22), and culture supernatant was collected 72 h post-treatment. The culture supernatant was centrifuged at 800 ϫ g (15 min), 1800 ϫ g (15 min), and 25,000 ϫ g (20 min), and the remaining supernatant was spun at 100,000 ϫ g (1 h) to pellet small apoptotic vesicles. Apoptotic vesicles were coated onto latex beads as described below and labeled with antibodies for analysis by flow cytometry or fluorescence microscopy.
Sucrose Density Gradient Centrifugation-For further purification of exosomes, the 100,000 ϫ g pellet was resuspended in 1 ml of 2.5 M sucrose, 20 mM Hepes/NaOH, pH 7.2. A linear sucrose gradient (2-0.25 M sucrose, 20 mM Hepes/ NaOH, pH 7.2) was layered on top of the exosomes suspension in a tube, and the sample was centrifuged at 100,000 ϫ g for 15 h. Gradient fractions (7 ϫ 1 ml) were collected from the top of the tube, diluted with 3 ml of PBS, and ultracentrifuged at 100,000 ϫ g for 1 h.
Caspase 3 Inhibitor Treatment and Staining of Apoptotic Cells-Cells were stained for annexin V as per manufacturer's instructions (Annexin V-FITC apoptosis Detection kit II, Calbiochem, WI). Briefly, cells were washed with binding buffer and stained with annexin V (1:40) for 10 min at room temperature. Cells were washed with binding buffer and stained with propidium iodide (1 g/ml) and analyzed by confocal microscopy. For caspase inhibition, the caspase-3-specific inhibitor Ac-DEVD-CHO (Calbiochem) was added to macrophages at a final concentration of 50 M 1 h before infection and maintained for the duration of assay (23)(24)(25).
Electron Microscopy-Exosome pellets were resuspended and fixed in phosphate buffer containing 2% glutaraldehyde and then loaded on Formvar/carbon-coated electron microscopy grids. The samples were contrasted in uranyl acetate and viewed with Hitachi H-600 transmission electron microscope.
Labeling with N-Rh-PE-The fluorescent phospholipids analog N-Rh-PE was inserted into the plasma membrane as previously described (26). Briefly, an appropriate amount of lipid was resuspended in ethanol. The ethanol solution was injected with a Hamilton syringe into serum-free Dulbecco's modified Eagle's medium while vigorously vortexing. This mixture was then added to the cells already infected with SmT M. avium 2151, and they were incubated for 60 min at 4°C. Subsequently, the medium was removed, and the cells were extensively washed with PBS. Labeled cells were fixed after 3 h of incubation at 37°C and immediately analyzed by confocal microscopy.
Analysis of Uninfected Bystander Cells-BMMs were infected with FITC-labeled SmT M. avium 2151, and, after 4 h of infection, an equal number of uninfected BMMs labeled with the cell tracker dye 7-amino-4-chloromethylcoumarin (10 M for 60 min) were added. Cells were examined by fluorescence microscopy for the presence of GPL in uninfected bystander cells.
Coupling of Exosomes or Apoptotic Vesicles to Latex Beads-The purified exosomes or apoptotic vesicles (30 g) were incubated with 4-m diameter aldehyde/sulfate latex beads (Interfacial Dynamics) for 15 min at room temperature. This was followed by the dilution with PBS, and binding reaction was continued for another 2 h. The reaction was stopped by addition of 100 mM glycine. Vesicle-coated beads were then washed three times in PBS and stained with specific antibodies.
Western Blot Analysis-For Western blots, equal concentration of protein from cell lysates or exosomes, as quantitated by the Micro BCA Protein Assay, were loaded on 10% SDS-PAGE gels, electrophoresed, and transferred onto polyvinylidene difluoride membrane (Millipore, Bedford, MA). The membranes were probed for iNOS (BD Transduction Laboratories) or total p38 (Cell Signaling) as described (16). Blots containing 10 g of exosome material were probed using antibodies against the M. tuberculosis LAM (1:500 dilution) or host hsp70 (R&D, 1/1000 dilution).
Macrophage Incubation with Exosomes or Apoptotic Vesicles-Thawed or fresh macrophages were cultured on non-tissue culture plates for 3-7 days and then re-plated at 1 ϫ 10 5 cells per well in 24-well tissue culture plate. The cells were allowed to adhere for 24 h prior to the addition of exosomes or apoptotic vesicles. The BMMs in culture media minus antibiotics were incubated with isolated vesicles for 24 h. The culture supernatant was harvested at indicated times for subsequent cytokine/ chemokine analysis.
ELISA-The levels of TNF-␣ and RANTES secreted into the culture medium by exosome-treated macrophages were measured using the ELISA kits from BD Pharmingen and eBioscience, respectively. Culture media collected from the macrophages were analyzed for cytokines according to manufacturer's instructions, and the cytokine concentrations were determined against standard curves.
Statistical Analysis-Data obtained from independent experiments were analyzed by a one-tailed Student t test. Differences were considered significant for p Ͻ 0.05.

RESULTS
Our prior studies established that GPLs are M. avium virulence factors (7,8). Moreover, GPLs are shown to accumulate inside the cell and have a long half-life inside a phagosome (11). To determine if GPLs can enter the endocytic pathway, we followed the trafficking of GPL inside infected cells using anti-GPL monoclonal antibodies. The specificity of the antibody was confirmed by TLC immunostaining comparing extracted lipids from SmT M. avium 2151, which express serotype 2 GPLs to an Rg 2151 isolate, which lacks GPL expression (28). As shown in Fig. 1A, we observed a single band only in lanes containing lipid extract from SmT M. avium 2151. This serotype-2 anti-GPL antibody was used to characterize the distribution of GPLs in SmT M. avium 2151-infected macrophages at 4, 24, 48, 72, and 96 h post infection. The fluorescence staining indicated that the GPLs were not only present on the mycobacterial surface (shown in merged images as yellow) but also separate from the mycobacteria (Fig. 1B). No cellular staining was observed with the anti-GPL monoclonal antibodies when the macrophages were infected with the GPLdeficient Rg M. avium 2151 (data not shown). The number of cells staining positive for GPLs was quantitated over time and indicated a gradual increase in the percentage of total GPL-positive macrophages between 4 and 96 h postinfection ( Fig. 1, C and D). Interestingly, this increase appeared to be due to enhanced staining of uninfected macrophages with the anti-GPL monoclonal antibodies suggesting transfer of GPLs from infected to uninfected cells (Fig. 1D). To further confirm this cell-to-cell transfer of GPLs, BMMs were infected with FITC-labeled SmT M. avium 2151. The infected cells were washed with PBS to remove any free mycobacteria prior to the addition of equal number of uninfected macrophages labeled with the fluorescent marker aminochloromethylcoumarin. As expected, by 24 and 48 h we observed a significant number of the uninfected, aminochloromethylcoumarin-stained cells to be positive for GPLs demonstrating the transfer of GPLs from infected to uninfected cells (Fig. 1E).
Analysis of SmT M. avium 2151-infected bone marrow-derived macrophages by fluorescence microscopy revealed significant release and trafficking of glycopeptidolipids. As noted above, mycobacteria are maintained within phagosomes that do not mature to phagolysosomes in non-activated macrophages. However, it is less clear whether released mycobacterial components can enter the late endosomal/lysosomal pathway or whether they can accumulate in specific compartments within infected cells. There are recent reports indicating that some mycobacterial lipids, including phosphoinositolmannoside (PIM) can enter distinct compartments in the endocytic pathway (14). Other mycobacterial lipids such as LAM show a more diffuse staining pattern in M. tuberculosis-infected macrophages (14). Previous studies looking at the trafficking of hydrazidelabeled mycobacterial lipids also reported these distinct membrane organelles carrying mycobacterial lipids, which were subsequently identified as MVBs (29).
To elucidate the route taken by GPL after release from mycobacterial phagosome and to characterize the GPL-positive compartment, we studied the colocalization of GPL with different endocytic markers, including LAMP1 (late endosomes), LAMP2 (late endosomes), and MHC class II (MIIC compartment). Studies with these markers showed significant colocalization with GPL ( Fig. 2A and data not shown), indicating that GPLs traffic to one of the late compartments within the BMMs endocytic pathway. Moreover, the GPL-positive compartments were found negative for the Golgi marker Grp78 ( Fig. 2A) and transferrin (data not shown). Based on these results and previous published studies, we hypothesized that the GPLs are trafficked predominately to the MVB. To further test our hypothesis, we looked for whether N-rhodamine-phosphatidylethanolamine (N-Rh-PE) colocalized with the GPL. N-Rh-PE when added to cells is efficiently internalized via endocytosis and has been shown to traffic specifically to MVBs in reticulocytes (30) and RAW 264.7 cells (31). N-Rh-PE added to the M. avium 2151infected cells showed marked colocalization with GPL (Fig. 2B). Together the data indicate that GPLs are released from M. avium-containing phagosomes, enter the endocytic pathway and localize, at least in part, to MVBs within infected cells.
MVBs can be trafficked toward the plasma membrane where they can fuse and release the intraluminal vesicles (ILVs) known as exosomes into the extracellular environment. However, this phenomenon of the transporting, docking, and fusion of MVBs with the plasma membrane is not well understood. Recent studies by Colombo and colleagues demonstrate the recruitment of Rab11 to MVBs and its role in the docking and fusion of MVB with the plasma membrane (31). As shown in  Rab11 and GPL was observed. This is in agreement with previously published data showing that Rab11 is involved in the trafficking of MVB toward the plasma membrane but not in the initial biogenesis of MVB (31). However, it is worth noting that many factors in addition to Rab11 play a role in the docking/ fusion of MVBs with the plasma membrane.
To determine whether infected macrophages release exosomes upon fusion of MVBs with the plasma membrane and whether the exosomes contain GPLs, we collected culture supernatant from M. avium 2151-infected J774 cells 3 days post infection. Parallel experiments were run with the uninfected J774 cells. Exosomes were purified from the culture supernatants by differential centrifugation and sucrose gradient as described (15). The purified exosomes were examined by whole mount transmission electron microscopy. Transmission electron microscopy analysis of exosomes prepared from M. avium-infected and uninfected J774 macrophages revealed homogenous populations of ϳ50 -100 nm vesicles (Fig. 4A). Exosome preparations appeared essentially devoid of any cell debris or chromatin material. We also tested other sucrose gradient fractions for vesicles by transmission electron microscopy (data not shown). However, no vesicles in the size range of exosomes were detected in the other isolated fractions.
To further assess our exosome preparation, the protein composition of the isolated vesicles was evaluated. The purified vesicles from M. avium-infected and uninfected J774 macrophages were coated on latex/aldehyde beads and probed for proteins known to be present on exosomes including LAMP1 and LAMP2 (lysosomal membrane proteins), MHCI and MHCII (antigen presentation), CD81 (tetraspannins), and CD86 (costimulatory molecule). As shown in Fig. 4B, the exosomes released from both uninfected and M. avium-infected cells were positive for LAMP1, LAMP2, MHCI, MHCII, CD86, and CD81 but lacked the Golgi marker grp78. The presence of hsp70, another marker exosomes was also confirmed by Western blotting (Fig. 4C). Also, as expected, the exosomes isolated from M. avium-infected but not from uninfected cells con-tained GPL (Fig. 4B). This further supports our hypothesis that GPLs are trafficked to MVBs and subsequently released extracellularly as part of exosomes. Therefore, this defines one plausible mechanism by which GPLs and possibly other mycobacterial components can be transferred from infected to uninfected cells. Indeed, a similar mechanism has been proposed by David Russell and colleagues for the transfer of PIM2, mycoside B, and other mycobacterial lipids to uninfected macrophages (14). We also found LAM to be present on the exosomes isolated from infected but not uninfected J774 cells (Fig. 4C). To test whether exosomes purified from M. avium 2151-infected macrophages can transfer GPLs to uninfected cells, we added 10 g of purified exosomes to BALB/c BMMs. The macrophages were fixed 24 h post-treatment and probed using the antibody against GPL serovar 2. As shown in Fig. 4D, serovar 2 GPLs could be detected in the macrophages treated with exosomes isolated from M. avium 2151-infected but not from uninfected BMMs.
To determine if the trafficking to MVBs and release on exosomes was specific to serovar 2 GPLs, we also studied the trafficking and release of GPLs from BMMs infected with M. avium serovar 1 and serovar 4 strains. Localization studies showed significant colocalization between N-Rh-PE and GPLs in macrophages infected with M. avium 101 (serovar 1) and M. avium A5 (serovar 4), indicating trafficking of these GPLs to MVBs (data not shown). We also added exosomes, purified from the culture supernatants of M. avium 101-and M. avium A5-infected J774 cells, to BMMs, and in both cases, GPL was detected in the exosome-treated cells (Fig. 4D). This suggests that the trafficking of GPLs to the MVBs and their release on exosomes is independent of GPL serotype.
These data above as well as work by David Russell and colleagues suggest that mycobacterial components, including GPL, LAM, PIM2, trehalsoe mono-and dimycolate, and phenolic glycolipids among others are released from phagosomes during a macrophage infection (14,29). Interestingly, some of these lipids have been shown to induce a proinflammatory response when introduced to uninfected macrophages. For example, PIM2 coated on microspheres could induce TNF-␣ and MCP-1 in interferon-␥ primed BMMs or thioglycollateelicited peritoneal macrophages (32). To determine if exosomes containing these mycobacterial lipids could induce a proinflammatory response, BALB/c BMMs were treated with exosomes, and the culture supernatants were assayed for cytokines and chemokines. As shown in Fig. 5 (A and B), BMMs treated with exosomes from infected cells induced increased levels of TNF-␣ and RANTES in a dose-dependent manner. This response was specific as exosomes isolated from uninfected macrophages stimulated little or no TNF-␣ or RANTES. TLR provides a means to recognize pathogen-associated molecular patterns and play a vital role in stimulating the host immune system (for review see Ref. 33). GPL, PIM2, and other mycobacterial glycolipids have been shown to stimulate macrophages through the TLRs (34,35). To investigate whether the exosome-mediated macrophage activation was dependent on TLRs and its associated adaptor molecule MyD88, we exposed C57BL/6 wild type (WT), BALB/c WT, TLR2Ϫ/Ϫ (C57BL/6 background), TLR4Ϫ/Ϫ (BALB/c background), and MyD88Ϫ/Ϫ (C57BL/6 background) BMMs to exosomes purified from uninfected and SmT M. avium 2151-infected J774 cells. Macrophage culture supernatants were collected 24 h post exosome treatment and assayed for TNF-␣. A significant decrease in TNF-␣ levels was observed in macrophages lacking TLR2 or TLR4 (data not shown). In addition, macrophages that lacked the MyD88 adaptor molecule were completely unresponsive to exosome treatment in the context of TNF-␣ production (Fig. 5C). Interestingly, the C57BL/6 BMMs appeared to be more responsive to exosome stimulation relative to BALB/c BMMs (Fig. 5, compare A to C).
Our data suggest that exosomes isolated from M. avium-infected J774 cells are proinflammatory. However, interpretation could be complicated by the potential contamination of our exosome preparation with apoptotic vesicles. Previous studies have shown M. tuberculosis-infected macrophages release apoptotic vesicles with antigen-stimulating activity (19,20). To evaluate whether apoptotic vesicle were involved in the observe response, we first tested whether apoptosis is occurring in the infected macrophage population by staining cells with annexin V, which stains surface-exposed phosphatidylserine, and propidium iodide, which is retained in late apoptotic cells (36). As shown in Fig. 6A, both M. avium-infected as well as uninfected J774 cells stained negative for both annexin V and propidium iodide. However, J774 cells induced to undergo apoptosis by serum starvation stained positive for both. The infection of the J774 cells was confirmed by LAM staining (data not shown). To further exclude the possibility that apoptotic vesicles are responsible for the proinflammatory activity, we treated infected macrophages with the caspase 3 inhibitor Ac-DEVD-CHO and isolated exosomes from these infected macrophages. The exosomes isolated from M. avium-infected cells showed similar TNF-␣ and iNOS production whether isolated from Ac-DEVD-CHO-treated or untreated J774 cells (Fig. 6, B and C). Moreover, apoptotic vesicles isolated form serum-starved J774 cells failed to induce TNF-␣ or iNOS production (Fig. 6, B and C). Finally, we observed minimal staining for Fc␥RII/III and annexin V on exosomes isolated from M. avium-infected J774 cells, similar to the levels seen on exosomes from uninfected cells. In contrast, apoptotic vesicles isolated from serum-starved J774 cells showed strong staining for both markers (Fig. 6D). Together, the data indicate that exosomes and not apoptotic vesicles are responsible for the proinflammatory activity.
The partially decrease in TNF-␣ production when using either TLR2-or TLR4-deficient BMMs suggest that the ability of exosomes isolated from M. avium-infected BMMs to stimulate a proinflammatory response is dependent on multiple mycobacterial components. To determine if GPLs is one of the major stimulatory molecules, we isolated exosomes from macrophages infected with GPL-sufficient SmT and GPLdeficient Rg M. avium 2151. Interestingly, the exosome isolated from Rg M. avium 2151-infected macrophages induced TNF-␣ production comparable to the levels induced in macrophages treated with exosomes from SmT M. avium 2151 (Fig. 7). This suggests that the GPLs present on exosomes are not functioning as major stimulators or that other mycobacterial components present on exosomes from Rg 2151-infected macrophages can compensate for the loss of GPLs.

DISCUSSION
The mycobacterial cell wall consists of a highly complex array of distinctive lipids, glycolipids, and proteins; many of which have been shown to have immune modulatory activity. Some of these modulatory functions can be attributed to the interaction of the mycobacteria surface components with the host macrophages during phagocytosis and likely involves pattern recognition receptors such as the TLRs. In addition, many mycobacterial components are also shed during the course of an infection. In vitro studies have shown that some of the shed mycobacterial components such as PIM when added to macrophages can induce TNF-␣ production (37). Trehalose dimycolate is another mycobacterial glycolipid released during an M. tuberculosis infection, which can promote macrophage activation when added to macrophages (38). Interestingly, these glycolipids as well as others have been shown to be transported from the mycobacteria phagosome to other cellular compartments and can be released extracellularly from infected macrophages (14). Our studies with M. avium and GPLs support the observation that mycobacterial glycolipids can be shed from phagosome and transported to other compartments within an infected cell. Moreover, we have found that GPLs are transported predominately to MVBs by 24 h post-infection with M. avium.
MVBs are endosomes that mature by gradually accumulating numerous small vesicles (ILVs) through invagination of the limiting membrane. In most cell types the MVBs are incorporated into the lysosomal compartment. However, in recent years several other functions have been attributed to MVBs, including the down-regulation of signal transduction by plasma membrane receptors (39). Interestingly, MVBs have also been shown to be an alternative secretory pathway initiated by the fusion of MVBs with the plasma membrane (40). This fusion  event leads to extracellular release of the ILVs as exosomes, which can be subsequently endocytosed by the neighboring cells (41).
The release of exosomes through fusion of MVBs with the plasma membrane occurs in many cell types (39,42). Proteins expressed on exosomes can vary between cell types, but there are some commonalities particularly among antigen-presenting cells (43). B lymphocyte-and DC-derived exosomes display abundant MHC class I and II molecules, co-stimulatory molecules CD80 and CD86, and adhesion molecule ICAM-1 (intercellular adhesion molecule 1) (15). The flow cytometric analysis of exosomes isolated from the M. avium-infected macrophages showed the presence of GPL along with the other markers known to be present on the exosomes, including MHC class I and II and CD86. Moreover, the immunofluorescence studies indicate that GPLs can be transferred from infected to uninfected BMMs through exosomes. These results further support our hypothesis that release of GPLs from an M. avium-infected BMM occurs at least in part through exosomes.
Previous studies by a number of research groups have shown that exosomes can function to modulate an immune responses, including immune stimulation and immune suppression (44,45). DC-derived exosomes also show potent capacity to generate anti-tumor immune responses in vivo (46). Tumor-derived exosomes, which are enriched in tumor antigens, are a novel source of antigens for promoting CTL cross-priming (47). Immature DCs secrete exosomes, which transfer functional MHC-peptide complexes to other DCs (48). Consistent with a stimulatory role in the immune response, exosomes secreted by DCs can elicit T-cell activation in vivo (49). Exosomes secreted by epithelial cells, despite their lack of co-stimulatory molecules, have also been shown to induce humoral immune responses (27).
Prior studies suggest that a number of other mycobacterial factors, including 19-kDa protein, mycolates, and phenol glycolipids, are present on small 50 -100 nm extracellular vesicles, which likely constitute exosomes (29). Although the immunogenic properties of purified mycobacterial lipids has been well documented, their role during an infection and how they may interact with cells when released from the mycobacteria is less understood. Studies by David Russell and colleagues have shown that some of the mycobacterial lipids present on these vesicles can elicit an immune response (32). However, these studies were carried out with lipids coated on microspheres and therefore do not reflected the natural route of their release. We initiated studies to define the immunogenicity of mycobacterial lipids in the context of a natural infection by isolating exosomes from mycobacteria-infected macrophages. Interestingly, our studies showed that exosomes isolated from infected cells induced significant TNF-␣ and RANTES production by BMMs.
Mycobacterial products released by the sequestered bacilli, including GPLs, LAM, PIM, and the 19-kDa lipoprotein can contribute to macrophage activation through pattern recognition receptors. The initial recognition of these mycobacterial pattern-associated molecular patterns by the innate immune system through TLR, and other pattern recognition receptors can contribute to triggering of the host immune response. Because exosomes are believed to contain pattern-associated molecular patterns, which can induce a proinflammatory response, it was possible that the observed macrophage response to exosomes was dependent on TLRs. To test this hypothesis, we treated TLR2 Ϫ/Ϫ , TLR4 Ϫ/Ϫ , or MyD88 Ϫ/Ϫ murine macrophages with exosomes derived from M. aviuminfected macrophages. The induction of proinflammatory response was found to be dependent on the TLRs as illustrated by the complete failure of MyD88-deficient macrophages to respond to exosomes.
Our results indicate that the mycobacterial components present on exosomes are released extracellularly and can interact with the neighboring cells through Toll-like receptors. Following exposure to the exosomes these uninfected cells respond by producing proinflammatory molecules like TNF-␣ and RANTES. This illustrates a previously undefined mechanism by which cells infected with intracellular pathogens can release pattern-associated molecular patterns to "alert" the immune system of an underlying infection and could function in conjunction with cytokines to promote a protective immune response. Importantly, we show that the proinflammatory response is not due to apoptotic vesicles, which have also been shown to contain immunogenic mycobacterial components (19).
Although our studies demonstrate a role for exosomes in promoting macrophage activation, we do not yet know the specific mycobacterial ligand(s) responsible for this response. Our studies with exosomes isolated from GPL-deficient Rg M. avium 2151 indicate that the lack of GPL on exosomes does not increase or decrease the stimulatory activity of the exosome. This suggested that either GPL does not play an important role in macrophage stimulation by exosomes or perhaps exosomes contain more than one ligand that is sufficient to induce macrophage activation. Therefore, the lack of GPL may be compensated by other mycobacterial factors such as PIMs, 19-kDa lipoprotein, or the mycolates. Future studies will be needed to address the mycobacterial pattern-associated molecular patterns present on exosomes that promote the proinflammatory response and their potential role in modulating the host response in vivo.
In summary, our studies demonstrate that GPLs are released from the mycobacteria-containing phagosome and are selec- tively trafficked to the MVBs. Consistent with the recently proposed role for MVBs, we found that exosomes carrying GPL are released from the mycobacteria-infected cells. Our studies show for the first time that exosomes isolated from mycobacteria-infected macrophages can induce a proinflammatory response and thereby modulate the overall immune response of the host.