Introduction Mycobacterium ulcerans is the etiologic agent of Buruli ulcer, a necrotic skin disease affecting humans living close to wetlands in tropical countries. The natural history and transmission of this mycobacteria are still obscure. Epidemiological studies suggest that swampy areas, and more specifically, aquatic environments, are the main ecosystems inhabited by M. ulcerans [1–7]. Many aquatic insects in this environment are predators that feed on herbivorous organisms, such as snails, which, after grazing on plants covered by M. ulcerans, act as passive hosts. It is conceivable that, following ingestion of prey contaminated with M. ulcerans, certain carnivorous aquatic insects might then transmit the bacteria to humans. We previously demonstrated that predatory aquatic insects, such as Naucoris cimicoides, ingest M. ulcerans–containing prey under laboratory conditions and are both hosts and vectors of this mycobacterial species [8–11]. Indeed, they are able to deliver invasive bacteria to laboratory mice, whose tails were exposed to the infected insects, with lesions developing 30 to 90 d later precisely at the point where the bite occurred. M. ulcerans is the only Mycobacterium species that localized within the salivary glands of these aquatic insects, where it can both survive and actively multiply without damaging insect tissues [8,10]. Furthermore, it has recently been shown that the lipid toxin mycolactone, the sole known virulence factor responsible for Buruli ulcer [12], is essential for the colonization of the salivary glands and that mycolactone-deficient mutants do not multiply in N. cimicoides [10]. Another striking feature of M. ulcerans is its ability to assemble into a biofilm, as first seen on the surface of aquatic plants [3]. Biofilms for human bacterial pathogens such as Pseudomonas aeruginosa, Haemophilus influenzae, and Vibrio cholerae have been well-studied [13] and consist of discrete bacteria surrounded by an extracellular matrix (ECM) [14,15]. Typically, the ECM shapes the bacterial network and can be crossed by channels, which play a critical role in water and nutrient circulation, as well as in interbacterial communication via quorum-sensing [16]. Biofilm formation confers a selective advantage for persistence under diverse environmental conditions and for resistance to antimicrobial agents, and also facilitates colonization of the host by the bacteria [13]. With regard to mycobacterial species, mutants of M. avium and M. smegmatis impaired in biofilm formation are less able to invade and translocate through bronchial epithelial cells and to form smegma in mice, respectively [17,18]. Molecular events involved in biofilm formation have already been reported in several studies undertaken on the genetically tractable M. smegmatis [19]. Recently, the GroEL1 chaperone was shown to be involved in mycolic acid biosynthesis during biofilm formation. Here, we show that the ECM of M. ulcerans differs from known biofilms since it is associated with only the outermost cell layer as opposed to classic biofilms in which all cells are surrounded by the matrix. Biochemical characterization of the ECM was performed and its role in pathogenesis at the different stages of the currently known life cycle investigated. Taken together, these findings provide insight into the factors that promote persistence in diverse environmental niches and infectivity of M. ulcerans to various hosts. Results M. ulcerans Colonies Are Covered by a Novel Type of ECM Harboring Vesicles Given the complexity of the life cycle of M. ulcerans, systematic examination was undertaken of the ultrastructure of the bacterium by scanning electron microscopy. Large clusters of M. ulcerans that were covered with a biofilm-like structure were detected in biopsy samples from patients with confirmed Buruli ulcers (Figure 1A). An analogous biofilm structure was also found in bacteria isolated from lesions from mice experimentally infected with M. ulcerans. According to the relative magnetic bead size, the surface of the recovered biofilm-like structure is estimated at about 200 μm × 50 μm, suggesting that one structure could harbor up to 105 bacteria. Further analysis of the same samples by Ziehl–Neelsen staining also revealed that, for each biofilm-like structure, there were less than ten free individual bacilli (unpublished data). A biofilm-like structure was also seen with M. ulcerans cultured in 7H9 broth, with or without Tween 80, and in 7H11 and 7H12 with or without PANTA (antimicrobial mixture) (unpublished data). The same amount of ECM was recovered from bacteria grown under all these culture conditions. Strains 1615 and 1G897 were examined at different time points, and bacilli-containing clusters were evident as early as day 10 (Figure 1B-1). From days 35 to 45, large cell aggregates measuring more than 100 μm were detected (Figure 1B-3). Higher-magnification micrographs revealed that entire clusters were surrounded by an abundant biofilm-like structure, the ECM (Figure 1B-4). All M. ulcerans clinical isolates that were cultured under the same conditions displayed the same biofilm-like structure, including the mycolactone-deficient mutant of 1615, mup045 (unpublished data). In contrast, other environmental or pathogenic mycobacteria, such as M. chelonei, M. fortuitum, M. kansasii, and M. tuberculosis, grew in vitro without displaying significant ECM despite their exhibiting large clusters of cells (Figure S1). M. marinum, the progenitor of M. ulcerans [20], formed discrete packets of cells that were quite distinct from the clusters, but displayed no ECM (Figure S1-3). These data show that the ECM is a peculiar feature of M. ulcerans colonies. Subsequent analysis by transmission electron microscopy (TEM) revealed that the ECM covers only the outermost bacterial layer, its thickness was estimated to range between 4 and 40 μm (Figure 2A, dotted circled area). Strikingly, very little matrix was found within the bacterial network (Figure 2B, arrows). This contrasts with classical biofilm, in which the bacteria are each individually surrounded by the matrix [14,15]. Furthermore, scanning microscopic pictures of M. ulcerans strains revealed the presence of vesicles on the surface of the ECM after 35 to 45 d of incubation (Figure 3A). The diameter of the vesicles varied between 50 and 200 nm (Figure 3B). Vesicles were isolated by ultracentrifugation from the wild-type bacterial ECM as well as from the mycolactone-deficient mutant (unpublished data). Together with ECM, vesicles could also be recovered from biopsies of mouse lesions by immunomagnetic separation (Figure 3C). Furthermore, the vesicles could be isolated independently from the ECM fraction by performing ultracentrifugation and were thus considered separately in the following analysis. To determine whether this matrix influences bacterial phenotype, comparison of the growth rate of M. ulcerans either harboring ECM, or from which ECM was carefully removed, was undertaken. No difference in colony-forming unit (CFU) counts of bacteria using Löwenstein–Jensen slants, and in metabolic activity using the Bactec radiometric method, was found with or without ECM removal from M. ulcerans (Figure 4), suggesting that biofilm does not confer a selective advantage for bacterial growth in vitro. In our culture conditions, the ECM re-forms in 2 wk. ECM Isolation and Effect of Its Removal on M. ulcerans Cultivability and Permeability The ECM fraction was isolated from broth-cultured M. ulcerans by mechanical disruption combined with Tween 80 detergent treatment, as typically used for other mycobacteria [21,22]. Fifteen seconds are sufficient for complete removal of ECM from bacteria (Figure 4A). We then compared the effect of the treatment on the cultivability of the treated bacteria. The same amount of CFUs was obtained for M. ulcerans with or without ECM (Figure 4B), showing that this mechanical disruption does not impair the cultivability of the bacteria. Furthermore, we checked whether this treatment modifies bacterial permeability. To this end, the level of potassium release by bacteria was monitored. No significant levels of potassium were released after mechanical vortexing for up to 60 s compared to untreated bacteria (Figure 4C). In addition, the presence of KatG, a catalase-peroxidase that is cytosolic or membrane-associated, was not detected by Western blot analysis in samples that had been treated for 15 s and 30 s (Figure 4D). Altogether, these data show that ECM can be efficiently isolated with a 15-s mechanical disruption, and this argues against major contamination of ECM by lysed bacteria. The ECM Protects Bacteria against Antimicrobial Agents To determine whether ECM plays a role in protecting bacteria from toxic compounds in the environment, the susceptibility of M. ulcerans, with and without ECM, toward chlorine and two common antibiotics was tested. This could be done since removal of ECM did not alter the growth of M. ulcerans, as shown above (Figure 4B). The minimum inhibitory concentration (MIC) of rifampin was 95) were considered as identified peptides. Proteins identified by a single peptide were validated by manual inspection of the MS/MS spectra. Human sera, ELISA, and Western blot analysis. The Buruli Patients group consisted of 30 patients recruited from the Centre de Diagnostic et de Traitement de l'Ulcère de Buruli in Pobè, Benin, and were included in a sero-epidemiological study, for which written consent had been obtained [9]. Nine out of 30 patients presented early clinical signs without ulceration (four nodules, two oedema, three plaques), ten patients with limited ulceration ( 10 cm). Diagnosis of M. ulcerans infection was by Ziehl–Neelsen staining of material taken from swabs of the lesions or directly from the biopsy for the early form and confirmed by PCR for M. ulcerans–specific IS2404 DNA [6]. The participants, who had given their written consent, were enrolled as volunteers in the study, the protocols of which were approved by the Ministry of Health in Benin. Serum was prepared from 8 ml of blood from each participant and tested for potential HCV and HIV exposure using Access HIV-1/2 automated immunoassay (MDA/98/58) and Sanofi Diagnostics Pasteur Access anti-HCV automated immunoassay (plus update on five other anti-HCV assays [MDA/96/26]). Proteins (10 μg) from ECM, bacterial lysate, membrane, and cytosolic fractions were coated onto 96-well Nunc Maxisorb plates by incubation overnight at 4 °C in 100 μl of PBS containing Tween 80 0.05%. The coated plates were then incubated with PBS containing 5% skimmed milk at room temperature for 2 h. After three washes in PBS/Tween 80, the samples were incubated for 1 h at 37 °C with human serum diluted 1:200 in PBS/Tween 80. After three further washes, plate-bound human immunoglobulins were detected using peroxidase-conjugated goat anti-human IgG (γ chain) antibodies (Sigma) and OPD (Dako, http://www.dako.com). The diluted sera were tested in triplicate and the average absorbance at 650 nm was expressed in optical density units. Proteins (60 μg) from lysates or ECM fractions were run in an SDS–polyacrylamide gel (4%–12%) (Bio-Rad), and the separated bands were transferred onto a 0.45-μm nitrocellulose membrane (Amersham). After blocking with 5% skimmed milk in PBS, the membrane was incubated with serum from humans diluted 1:100 in PBS containing Tween 80 0.05% for 90 min at 37 °C. After two washes with PBS containing Tween 80, anti-human IgG (γ chain) peroxidase-conjugated antibodies (Sigma) at 1:2,000 and 0.5 μg/ml DAB (Interchim) was used, respectively, to detect human IgGs bound to the different bands. Analysis by fluorescence microscopy of surface carbohydrates of M. ulcerans ECM. Bacteria were fixed in 2.5% (w/v) formaldehyde in PBS buffer and surface carbohydrates labeled with Texas red hydrazide (Molecular Probes, http://probes.invitrogen.com) or with calcofluor white M2R (Sigma). Bacteria were then stained by DAPI, and labeled carbohydrates were visualized directly using a Zeiss Axioskop 20 fluorescence microscope and the AxioVisionLE 4.2 SP1 program (http://www.zeiss.com) used to perform the 3-D reconstruction. Analysis of lipid and mycolactone contents of ECM. ECM obtained by treating bacteria with Tween 80 0.05% and glass beads was extracted by phase partitioning. The aqueous layer from the H2O/CHCl3/CH3OH partition was concentrated, the polymers precipitated overnight at 4 °C with six volumes of cold ethanol, and the precipitates collected by centrifugation at 14,000g for 1 h. The lipoarabinomannan content of this fraction was analyzed by SDS-PAGE and immunoblotting using the CS-35 antibody as described [49]. The interphase derived from the partition experiment was extracted three times with water before precipitating the polymers with ethanol and submitting them to acid hydrolysis in 2 M trifluoroacetic acid for 2 h at 120 °C. The monosaccharide constituents of this fraction were then analyzed by TLC, and their migration profile was compared to that of known standards. Total lipids from bacterial cells treated with Tween 80 0.05%, or untreated, were extracted as described [50] and analyzed by TLC on silica gel 60–precoated plates F254 (Merck, http://www.merck.de). Extraction of mycolactone and cytotoxicity tests was performed according to George et al. [40] either using the bacterial pellet (cells without ECM) or ECM [13]. The carbohydrate content of the ECM material was measured by a colorimetric method [51]. Cytotoxicity assays. Bone marrow–derived macrophages were obtained by seeding 105 bone marrow cells from 8-wk-old C57BL/6 mice per well in RPMI 1640 supplemented with 10% heat-inactivated fetal calf serum and 10% L-cell-conditioned medium. Culture medium was changed at day 4 and just before adding mycolactone at day 7. HeLa cells and Cos cells (American Type Culture Collection, http://www.atcc.org) were cultured in Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated fetal calf serum. Proliferating cells were seeded in 96-well microtitration plates at a density of 105 cells/well, which were further incubated for 24 h at 37 °C under 5% CO2 in air before each assay. Various concentrations of vesicles or mycolactone in ethanol were added (2 μl/well). The mycolactone used as reference was purified as previously reported [12]. After 24 h incubation in the above conditions, cytotoxicity was then assessed by addition of 20 μl of dimethylthiazolyl diphenyl tetrazolium bromide solution (MTT, Sigma) (7.5 mg/mL) to each well and further incubated for 4 h at 37 °C to allow the formation of formazan. Formazan crystals were then dissolved with 100 μl of 10 % SDS in 10 mM HCl. The optical density of each well was measured at 595 nm using a Multiwell plate reader. The values given are the average of two replicates and are representative of four independent experiments. The 50% inhibition concentration was determined by curve fitting. Quantification of M. ulcerans DNA in insect hemolymph using real-time PCR. Adult N. cimicoides water bugs were collected and housed as described previously, then fed with grubs of Phormia terrae novae (Verminière de l'Ouest, http://www.verminieredelouest.fr) that had been inoculated beforehand with M. ulcerans, with or without ECM, in 30 μl by using a 25-gauge needle. Six hours after feeding, the insect hemolymph was collected with an insulin syringe [10]. Pooled hemolymph (100 μl) was added to 100 μl of cold distilled water. The samples were washed three times by centrifugation (14,000g for 15 min) in distilled water and resuspended in 50 μl of 50 mM NaOH and heated at 95 °C for 15 min. Real-time PCR was performed using brilliant SybrGreen Q PCR mix (Stratagene, http://www.stratagene.com) containing Taq polymerase, 2.5 mM MgCl2, 100 μM (each) deoxynucleoside triphosphate and 20 pM primers. The primers were MLF (5'- CCCTTCGACGTCATCAAGAAA −3′) and MLR (5'- CCGACTGACCGATGAGCAA −3′), leading to amplification of a 63-bp region of the mls genes [52]. After 15 min at 95 °C, the DNA was amplified by 30 cycles of 45 s at 95 °C; 1 min at 61 °C, and 45 s of elongation at 72 °C on an MX3000P apparatus (Stratagene). The dissociation curve was performed between 55 °C and 95 °C. Susceptibility of M. ulcerans to antibiotics and chlorine. The MIC of rifampin and amikacin, inhibiting >99% of the bacteria, was determined as previously described [53]. To measure chlorine susceptibility, ∼108 bacteria, with or without ECM, were suspended in solutions containing a range of chlorine concentrations (20–200 mg per liter). After 60 min at 25 °C, residual chlorine was neutralized with sodium thiosulfate [54] and bacterial viability determined by inoculation onto Löwenstein–Jensen slants. Virulence assays in mice. Suspensions (30 μl) containing 5 × 103 bacteria, with or without ECM, were injected subcutaneously into the tail of ten female Balb/c mice (Charles River Laboratories, http://www.criver.com). Mice tails were examined weekly over 6 mo. Statistics. The non-parametric Mann-Whitney U test was used for statistics. A p-value of < 0.05 was considered significant. Supporting Information Figure S1 Scanning Electron Micrographs of Four Mycobacterial Species at the End of Exponential Growth in 7H9 Culture Medium (1) M. chelonei; (2) M. kansasii; (3) M. marinum; (4) M. tuberculosis. Scale bar: 5 μm. (8.7 JPG) Click here for additional data file. Table S1 Detailed Annotation of Proteins Localized in Vesicles of M. ulcerans (117 KB DOC) Click here for additional data file.