Identification of Phosphatidylinositol Mannoside as a Mycobacterial ...

INFECTION AND IMMUNITY, Sept. 1997, p. 3896–3905 0019-9567/97/$04.0010 Copyright © 1997, American Society for Microbiology

Vol. 65, No. 9

Identification of Phosphatidylinositol Mannoside as a Mycobacterial Adhesin Mediating Both Direct and Opsonic Binding to Nonphagocytic Mammalian Cells HEINRICH C. HOPPE,1 BAREND J. M. DE WET,1 COLETTE CYWES,1 ´ ,2 AND MARIO R. W. EHLERS1* MAMADOU DAFFE Department of Medical Biochemistry, University of Cape Town Medical School, Observatory 7925, Cape Town, South Africa,1 and Institut de Pharmacologie et de Biologie Structurale du Centre National de la Recherche Scientifique and Universite´ Paul Sabatier, 31062 Toulouse Cedex, France2 Received 28 March 1997/Returned for modification 14 May 1997/Accepted 13 June 1997

The molecular basis for the binding of Mycobacterium tuberculosis to nonphagocytic cells, which are readily infected in vitro, and the in vivo significance of this interaction are incompletely understood. Of six cell types tested, we found that only two, Chinese hamster ovary (CHO) fibroblasts and primary porcine aortic endothelial cells, were able to bind M. tuberculosis H37Rv efficiently in vitro. Binding to both CHO and endothelial cells was markedly (three- to fivefold) enhanced by 10 to 20% human or bovine serum, suggesting that the bacteria were coated by a serum opsonin. Preincubation with individual candidate opsonins revealed that recombinant human mannose-binding protein (rMBP), fibronectin, and transferrin were each able to enhance binding threefold. Preincubation of bacteria in serum depleted of mannan-binding lectins or in genetic MBP-deficient serum resulted in enhancements that were only ;60 and 58%, respectively, of that produced by preincubation in control serum. In contrast, serum depleted of fibronectin or transferrin retained its opsonizing capacity, suggesting that the latter two are not significant opsonins in whole serum. Binding of M. tuberculosis and Mycobacterium smegmatis to both CHO and endothelial cells in the presence or absence of serum was blocked (60 to 70%) by a monoclonal antibody, MAb 1D1, selected for recognition of intact bacilli. The 1D1 antigen was purified from mycobacterial cell walls and chemically identified as a polar phosphatidylinositol mannoside (PIM). Latex beads coated with purified 1D1 antigen bound to CHO cells, which was enhanced threefold by serum and abolished by periodate treatment, suggesting a requirement for the PIM mannoses in opsonic adhesion. This was likely mediated, at least in part, by serum MBP, as rMBP bound strongly to 1D1 antigen in both thin-layer chromatography overlay and plate binding assays, the latter in a mannan-inhibitable manner. This is the first demonstration that mycobacterial PIMs can function as adhesins for binding to nonphagocytic cells, both directly and after opsonization with serum proteins, including MBP. It is estimated that Mycobacterium tuberculosis infects onethird of the world’s population. This organism is a facultative intracellular pathogen, and the establishment of a primary infection is thought to be critically dependent on the colonization of host mononuclear phagocytes (7, 17). It is unclear to what extent M. tuberculosis invades nonphagocytic cells in vivo and whether this plays a role in the pathogenesis of tuberculosis, but indirect evidence suggests that invasion of nonphagocytic cells may be a feature of specific stages of the disease (9, 34). However, it is well established that M. tuberculosis readily infects nonphagocytic cells in culture, although the mechanisms involved are largely unknown (2–5, 18, 31, 32, 39, 44). In contrast, the interaction of M. tuberculosis with mononuclear phagocytes has recently received considerable attention, including efforts to delineate the molecular basis for the adherence of bacilli to host cells. M. tuberculosis binds to monocyte/macrophage complement receptor type 1 (CR1), CR3, and CR4 (20, 41, 45). Although this interaction is enhanced by opsonization with complement component C3 and its activation products, there is evidence that the bacteria are also able to bind CR3 in a direct, nonopsonic manner (11, 40, 45). M. tuberculosis also binds nonop-

sonically to the macrophage mannose receptor, presumably via surface-expressed mannosylated glycoconjugates that may include lipoarabinomannan (LAM) (40), although the exact identities of surface-expressed M. tuberculosis ligands for the mannose receptor remain to be established (35, 46). Additionally, it has been shown recently that binding of M. tuberculosis to human alveolar macrophages is significantly enhanced by opsonization with surfactant protein A (SP-A) (13). SP-A is a member of the collectin family, which consists of oligomeric proteins comprised of subunits that each contain an N-terminal collagen-like domain and a C-terminal carbohydrate recognition domain; this family also includes human mannosebinding protein (MBP) and bovine conglutinin (14). Molecular data for the binding of M. tuberculosis to nonphagocytic cells are sparse, although some inferences can be made from studies with related mycobacteria. Mycobacterium bovis BCG attaches to and invades T24 bladder tumor cells, in a process that involves opsonization with fibronectin (FN) and binding to the a5b1 integrin receptor (26). Mycobacteria express a number of FN-binding proteins (reviewed in reference 43), some of which may mediate FN-dependent attachment to nonphagocytic cells. This has been shown for the 50-kDa FN attachment protein (FAP), which is expressed by several mycobacteria, including M. tuberculosis, and which mediates FNdependent invasion of nonphagocytic host cells by both Mycobacterium leprae and M. bovis BCG (43). A similar role for the

* Corresponding author. Phone: (27-21) 406-6335. Fax: (27-21) 477669. E-mail: [email protected]. 3896

VOL. 65, 1997

PIM AS A MYCOBACTERIAL ADHESIN

M. tuberculosis FAP homolog (28) in the binding and invasion of nonphagocytic cells has yet to be demonstrated. We are attempting to dissect the molecular details of the binding of M. tuberculosis to nonphagocytic cells, because these data may contribute to a more comprehensive understanding of the interaction of this pathogen with host cells. We have examined the binding of M. tuberculosis H37Rv to several mammalian cell lines in culture and found that binding was restricted to certain cell types. Binding was markedly enhanced by serum opsonins, notably MBP. Direct and serum-enhanced binding was mediated by a cell surface glycophospholipid, identified as a polar phosphatidylinositol mannoside (PIM), which strongly bound MBP in vitro. These data establish a role for surface-exposed polar PIMs as M. tuberculosis adhesins that mediate attachment to nonphagocytic cells, both directly and after coating with serum opsonins, which likely include MBP. MATERIALS AND METHODS Buffers and reagents. Mannan-agarose, polystyrene latex beads (1.16-mm diameter), mannan (from Saccharomyces cerevisiae), D-mannose, D-galactosamine, N-acetyl-D-glucosamine, orcinol, goat anti-human FN and anti-human transferrin immunoglobulin (Ig), and goat IgG were from Sigma Chemical Co. (St. Louis, Mo.). Human plasma FN, human serum transferrin (partially iron saturated), peroxidase-conjugated sheep anti-mouse Ig, and the ISOStrip isotyping kit were from Boehringer Mannheim (Mannheim, Germany). Silica Gel 60 thin-layer chromatography (TLC) plates were from Merck (Darmstadt, Germany), silica gel (32 to 63 mm) for column chromatography was from Saarchem (Krugersdorp, South Africa), and poly(isobutylmethacrylate) was from Aldrich Chemical Co. (Gillingham-Dorset, England). Human serum was freshly prepared from the blood of healthy volunteers. Vascular endothelial cells were obtained from the Department of Cardiothoracic Surgery, Groote Schuur Hospital (Cape Town, South Africa). Recombinant human MBP (rMBP) and MBP-deficient human serum ([MBP] , 50 ng/ml) from an individual with genetic MBP deficiency were kindly donated by R. Alan B. Ezekowitz (Children’s Hospital, Harvard Medical School, Boston, Mass.). Phosphate-buffered saline (PBS) consisted of 0.14 M NaCl, 8 mM Na2HPO4, 2.7 mM KCl, and 1.4 mM KH2PO4 (pH 7.4), and Tris-buffered saline (TBS) consisted of 0.15 M NaCl and 0.02 M Tris (pH 7.5). Bacterial strains. M. tuberculosis H37Rv (ATCC 27294) and Erdman (ATCC 35801) and Mycobacterium smegmatis (ATCC 19420) were cultured as described previously (11). Mycobacterial stocks prepared for the binding assays (see below) were quantitated as follows. Bacteria were heat killed, syringed repeatedly to disperse clumps, fixed in 2.5% glutaraldehyde, diluted in PBS, and mixed with an equal volume of a Candida albicans reference standard. The latter was prepared by suspending glutaraldehyde-fixed C. albicans in PBS and counting the yeast cells in a Neubauer chamber. The bacterium-yeast mixture was stained with acridine orange (1.4 mg in 10 ml of PBS, 30 s), pelleted and washed three times, mounted in PBS under a glass coverslip, and quantitated under fluorescence microscopy by counting particles in several optical fields. The concentration of mycobacteria in the original sample was derived from the C. albicans/mycobacteria ratio. Mammalian cells. Cell types used in this study were CHO-K1 (Chinese hamster ovary fibroblasts; ATCC CCL-61), MCF-12F (human mammary gland epithelial cells; ATCC CRL-10783), T-47D (epithelial-like human breast ductal carcinoma; ATCC HTB-133), WI-38 (human embryonic lung fibroblasts; ATCC CCL-75), HeLa (human cervix carcinoma; ATCC CCL-2), and primary porcine vascular endothelial cells. CHO and MCF-12F cell stocks were maintained in Dulbecco modified Eagle medium (DMEM)–Ham’s F12 (1:1 mixture) supplemented with 10% fetal bovine serum (FBS) and antibiotics (penicillin and streptomycin). T-47D, WI-38, and HeLa cells were maintained in DMEM containing 10% FBS and antibiotics. Endothelial cells were released from freshly dissected porcine aorta by collagenase treatment and plated onto FN-coated surfaces in M199 medium with Earle’s salts and L-glutamine, supplemented with 10% FBS, heparin, endothelial growth supplement, and gentamicin, as described previously (50). M. tuberculosis binding assays. Mammalian cells were plated onto 12-mmdiameter glass coverslips, incubated with bacteria, fixed, stained, and quantitated exactly as described previously (11), except that cells were grown for 48 h before infection and incubations with bacteria were for 6 h at a multiplicity of infection (MOI) of 100:1. Results are expressed as the percentage of cells associated with one of more bacteria, given as the mean of triplicate wells 6 standard deviation, and the significance of differences was determined by Student’s t test. Partial isolation of bovine serum lectins. One hundred milliliters of FBS was diluted with an equal volume of TBS containing 40 mM CaCl2 and 0.02% sodium azide and applied to a 5-ml column of mannan-agarose at 0.5 ml/min. The column was washed with 40 ml of TBS–10 mM CaCl2 to remove unbound material, and mannan-binding lectins were eluted with 30 ml of TBS containing

3897

TABLE 1. Abilities of M. tuberculosis cell wall-specific MAbs to inhibit binding to CHO cellsa MAb

Heavy-chain isotype

Antigen (kDa)b

% Inhibition of M. tuberculosis binding

4F11 2B2 4F5 2A8 4B8 4G9 2B10 1D1 3D6 4H12 1A2

IgG2a IgG2a IgG1 IgG1 IgM IgM IgM IgM NDd ND ND

32 75 85 65 (GroEl) NSc NS NS 12–16 12–16 NS NS

0 0 0 0 2 0 17 6 16 63 6 7 46 6 5 0 0

a A panel of MAbs was raised against M. tuberculosis H37Rv cell wall extracts and selected for recognition of intact, whole bacteria. CHO cells were incubated with M. tuberculosis in undiluted MAb hybridoma supernatants. Percent inhibition of binding was calculated with reference to a control containing medium without MAbs; values are means 6 standard deviations derived from triplicates of a single experiment. b Determined by Western blotting. c NS, no signal. d ND, not determined.

10 mM EDTA and 0.02% sodium azide. The Ca21 concentration in the eluate was adjusted to 20 mM with 0.2 M CaCl2; the eluate was subsequently dialyzed overnight against TBS–10 mM CaCl2, whereafter it was concentrated to 1.5 ml by centrifugal ultrafiltration in Ultrafree filtration units (Millipore, Bedford, Mass.). Opsonin-depleted serum. Serum depleted of mannose-binding lectins, FN, or transferrin was prepared by incubating 2 ml of 20% human serum in PBS for 18 h with ;2 ml of, respectively, mannan-agarose, anti-FN-agarose, or anti-transferrin-agarose (the latter two were generated by coupling goat anti-human FN and anti-human transferrin IgG to CNBr-activated agarose). Appropriate control sera were prepared similarly by incubating serum with agarose alone or with agarose coupled to nonspecific goat IgG. Bacteria were preincubated for 1 h with these opsonin-depleted sera and then added to CHO cells in serum-free medium. MAbs. Two rounds of monoclonal antibody (MAb) production were carried out. In the first round, BALB/c mice were immunized intraperitoneally with 2 3 8 10 irradiated M. tuberculosis H37Rv bacilli suspended in Freund’s incomplete adjuvant, and MAb production was carried out according to standard protocols. Hybridoma culture supernatants were tested for the presence of anti-M. tuberculosis MAbs by an enzyme-linked immunosorbent assay (ELISA). The solid phase of the ELISA method consisted of whole, irradiated M. tuberculosis H37Rv dried onto microtiter plate well surfaces (approximately 4 3 106 bacteria per well) and fixed in 80% methanol. This approach yielded 12 MAbs, all of which recognized the M. tuberculosis hsp65 GroEl-homologous heat shock protein (49), as determined by Western blotting. For the second round of MAb production, mice were immunized with an hsp65-depleted M. tuberculosis extract (containing approximately 0.5 mg of protein) emulsified in Freund’s incomplete adjuvant. The extract was prepared by sonicating M. tuberculosis H37Rv bacilli in 2% sodium dodecyl sulfate (SDS) for 4 min, followed by boiling for 5 min. Triton X-100 was added to a final concentration of 4% (vol/vol), and the extract was dialyzed against PBS. The extract was depleted of hsp65 by applying it to an hsp65 MAb-agarose column, which had been prepared by preincubating protein A-agarose with anti-hsp65 MAb-containing hybridoma supernatants; the M. tuberculosis extract was then eluted with PBS. The absence of hsp65 from the depleted extract was confirmed by Western blotting with anti-hsp65 MAbs. MAbs prepared with the extract were analyzed by Western blotting, and the heavy- and light-chain isotypes were determined with an ISOStrip kit (Table 1). Purification of 1D1 Ag. The 1D1 antigen (1D1 Ag) (which is recognized by MAb 1D1; see Results) was purified from bulk cultures of M. tuberculosis H37Rv or M. smegmatis. Sedimented bacteria were lyophilized, and the pellet was extracted three times with chloroform-methanol (2:1) at 50°C for 1 h. The lipid extract was removed, and the delipidated bacteria were boiled for 2 h in 70% ethanol under reflux. The remaining insoluble cell wall fraction was further extracted by boiling for 10 min in 2% SDS. The 70% ethanol extract was concentrated by evaporation and further fractionated, either by reversed-phase high-performance liquid chromatography (HPLC) or by silica gel column chromatography. HPLC was performed on a C-4 column, using a 10 to 100% acetonitrile gradient and a flow rate of 0.7 ml/min (total run time, 75 min). HPLC fractions were collected at 2-min intervals and tested for 1D1 Ag by ELISA, using 1D1 hybridoma supernatant as the primary antibody and peroxidaseconjugated sheep anti-mouse Ig (1:1,000) as the secondary antibody. Alterna-

3898

HOPPE ET AL.

INFECT. IMMUN. Deacylation of 1D1 Ag. Purified antigen (50 mg of carbohydrate, as determined by phenol-sulfuric acid assay [42]) was treated with 250 ml 0.1 M NaOH at 37°C for 2 h and neutralized with acetic acid, and fatty acids were extracted with 100 ml of chloroform (23). The aqueous residue was considered lipid free and was used as deacylated 1D1 Ag in bacterial binding assays as described in Results.

RESULTS

FIG. 1. Binding of M. tuberculosis H37Rv to mammalian cells in culture. Cells were incubated with bacteria (MOI of 100:1) for 6 h in medium containing 20% human serum, and the percentage of cells that bound one or more bacteria was determined by counting under fluorescence microscopy after staining with acridine orange.

tively, the 70% ethanol extract was fractionated by silica gel column chromatography, developed with a chloroform-ethanol-water (55:45:10) solvent mixture. Eluted fractions were assayed for 1D1 Ag by ELISA as described above. Analysis of 1D1 Ag. Bacterial extracts and HPLC fractions were analyzed by SDS-polyacrylamide gel electrophoresis (PAGE) and Western blotting. Samples were run on SDS–15% polyacrylamide gels and stained by the periodic acidsilver nitrate method (48) or transblotted onto nitrocellulose for reaction with MAb 1D1, using standard Western blotting protocols. Alternatively, the composition and purity of positive fractions was assessed by TLC on silica gel plates (solvent mixture as described above). The primary stain was an orcinol spray reagent to detect hexose-containing compounds (16); other stains were used as appropriate. TLC-resolved fractions were analyzed for immunological reactivity by a TLC overlay assay as follows. TLC plates, dried at 50°C, were immersed in a solution of 0.01% poly(isobutylmethacrylate) in hexane for 30 s and air dried. The plates were then treated with blocking solution (1% skim milk and 0.1% Tween 20 in TBS) for 1 h, incubated in MAb 1D1 hybridoma supernatant (1 h), washed, incubated with peroxidase-conjugated sheep anti-mouse Ig antiserum (in TBS; 1 h), and finally stained with the peroxidase substrate 4-chloro-1-naphthol. Chemical analyses of 1D1 Ag were performed as follows. A sample (0.5 mg) of the purified antigen was suspended in 1 ml of methanol-HCl (1.5 N), and the mixture was incubated at 80°C for 16 h. After evaporation under nitrogen, an equal volume of water and petroleum ether (boiling point, 50°C) was added to the methanolyzed products. The organic phase was dried under nitrogen; the polyols were trimethylsilylated and then analyzed by gas chromatography (GC). The organic phase that contained fatty acid methyl esters was also analyzed by GC. Authentic samples of various fatty acid methyl esters and trimethylsilylated sugars were separately analyzed by GC. Cochromatography of these standard compounds with the methanolyzed products allowed the identification of the different antigen constituents. GC was performed on a Girdel G30 apparatus equipped with a fused-silica capillary column (25 m by 0.22 mm) coated with OV-1 (0.3-mm film thickness; Spinal). A temperature gradient from 100 to 280°C (2°C/min) was used. MBP-binding assays. A modified TLC overlay assay was developed in which the primary antibody was replaced by biotinylated rMBP and streptavidin-peroxidase was used in place of the secondary antibody. rMBP was biotinylated by incubation with N-hydroxysuccinimide-biotin in PBS for 4 h at room temperature followed by dialysis to remove unreacted N-hydroxysuccinimide-biotin; biotinylated MBP was used at 10 mg/ml in the overlay assay. Additionally, binding of rMBP to purified 1D1 Ag, mannosylated LAM (Man-LAM), and Escherichia coli lipopolysaccharide (LPS) was quantitated in a modified ELISA. Each microtiter plate well was coated with approximately 1 mg of test compound (1D1 Ag concentration was estimated by the phenol-sulfuric acid carbohydrate assay [42]). Wells were blocked with 1% BSA in PBS, incubated with biotinylated rMPB (10 mg/ml in blocking buffer), washed, and then incubated with streptavidin-peroxidase. After reaction with the chromogenic substrate o-phenylenediamine, bound peroxidase was quantitated spectrophotometrically at 492 nm. Binding of coated latex beads to CHO cells. Four milliliters of pooled 1D1 Ag-containing HPLC fractions was evaporated in glass vials to remove acetonitrile and incubated with 1.4 3 108 latex beads for 2 h at room temperature with agitation. The beads were then pelleted and incubated in PBS containing 5% BSA for 2 h. Alternatively, antigen-coated beads were incubated for an additional 2 h in the dark in 0.2 M sodium acetate (pH 4.5) containing 0.1 M sodium periodate (38) and then washed in PBS before being blocked in 5% BSA. Control beads were incubated only in 5% BSA. Beads were subsequently added to CHO cells plated onto glass coverslips in 24-well plates containing DMEM or DMEM supplemented with 20% FBS (1.6 3 106 beads added per well); after incubation for 6 h at 37°C, the cells were washed in PBS and fixed in 2.5% glutaraldehyde. Binding was quantitated by counting the number of beads per cell under light microscopy.

Binding of M. tuberculosis to mammalian cells. Six mammalian cell types were tested for the ability to bind M. tuberculosis H37Rv. M. tuberculosis bound strongly to CHO and porcine endothelial cells, with 28% 6 4% and 35% 6 7%, respectively, of the cells bound to one or more bacteria in a representative experiment (Fig. 1). The levels of binding were found to vary from experiment to experiment, but generally between 20 and 40% of CHO cells bound one or more bacteria; in any particular experiment, the level of binding varied by ,20% between triplicates. Interestingly, binding was typically uneven across the CHO cell monolayer, such that only up to 40% of cells bound bacteria, with an average of five bacteria bound per cell, although some cells were more heavily infected (Fig. 2). At an MOI of 100:1, with an average of 30% of cells infected, and with five bacteria per cell, the percentage of the bacterial inoculum that bound to cells was ;1.5%. The reasons for the variability in binding between experiments are not clear but may relate to cell cycle-dependent variations in receptor expression. Importantly, although the absolute binding values varied, values relative to controls remained constant. For this reason, care was taken to ensure that in subsequent experiments, in which the effects of various factors on binding were examined (see below), all necessary controls were included in every experiment. In contrast, HeLa and MCF-12F cells did not bind bacteria, and only modest binding was exhibited by WI-38 and T-47D cells (5% 6 2% and 4% 6 3%, respectively, of the cells bound bacteria) (Fig. 1). Poor binding to HeLa cells was also ob-

FIG. 2. Fluorescence micrograph of CHO cells heavily infected with M. tuberculosis H37Rv. Magnification, 3950; bar, 20 mm (inner length of rectangle).

VOL. 65, 1997

FIG. 3. Serum dependence of binding of M. tuberculosis H37Rv to CHO cells (A) and porcine aortic endothelial cells (B). Bacteria were preincubated in 20% fresh or heat-inactivated human or bovine serum, or in PBS alone, and then incubated with cells in serum-free medium, stained, and counted as described in the legend to Fig. 1. p, P , 0.01 compared to PBS (no serum) controls. (C) Opsonizing activity of FBS fractions eluted from a gel filtration column. Bacteria were preincubated in each fraction and then incubated with CHO cells in serumfree medium. Cell-binding data are shown below each lane of the corresponding fraction run on an SDS–7.5% polyacrylamide gel; percent binding after preincubation in 10% FBS or PBS alone is also shown.

served with M. tuberculosis Erdman. Moreover, extending the duration of incubation of cells with M. tuberculosis from 6 to 18 h did not significantly affect the results (data not shown). Opsonization with serum enhanced the binding of M. tuberculosis H37Rv to CHO and endothelial cells. Preincubation in human or bovine serum enhanced the binding of M. tuberculosis to CHO and endothelial cells three- to fivefold (Fig. 3A and B), suggesting that serum components were opsonizing the bacteria for binding to the cells. This pattern of binding was always observed with the H37Rv strain (experiments were performed on at least 20 separate occasions). A similar pattern was seen with M. smegmatis. In contrast, the Erdman strain bound to CHO cells in a serum-independent manner (discussed further below; see also Fig. 6). The relatively inefficient binding of M. tuberculosis H37Rv to T-47D cells was also serum dependent, whereas binding to WI-38 cells was serum independent (results not shown). Subsequent experiments were performed predominantly with CHO cells (except where noted), because of ease of culture and no significant loss of viability or detachment during the in vitro binding assays. Identification of candidate serum opsonins. FBS was separated into 12 fractions by gel filtration (Sephacryl S-200, eluted with PBS), and each fraction was assayed for its opsonin content by incubation with M. tuberculosis H37Rv, after which the bacteria were allowed to bind to CHO cells in serum-free medium. The protein composition of the fractions was determined by nonreducing SDS-PAGE (Fig. 3C). Of the 12 fractions, the first 7 were able to enhance M. tuberculosis binding to CHO cells 2.5- to 5-fold (Fig. 3C). These seven fractions differed markedly in protein composition, which strongly suggested that several serum components can act as opsonins for M. tuberculosis. Similar results were obtained with human serum.


3899

The abilities of various proteins to enhance the binding of M. tuberculosis H37Rv to CHO cells under serum-free conditions were tested (Fig. 4). Preincubation in 20 mg of rMBP per ml resulted in a threefold increase in bacterial binding. Preincubation in FN or transferrin (both at 0.5 mg/ml) also increased bacterial adhesion approximately threefold, and in each case this was similar to the effect of 20% FBS (Fig. 4A). The ability of serum lectins to opsonize M. tuberculosis was analyzed further by testing a crude lectin isolate from FBS (see Materials and Methods). Bacteria were preincubated in PBS, in PBS plus 20% FBS, or in the lectin isolate and were subsequently incubated with CHO cells in serum-free medium. Preincubation in the crude lectin isolate enhanced bacterial binding to an extent similar to that found with 20% FBS or rMBP (Fig. 4B). The increase in binding was inhibited by 52% when mannan (1 mg/ml) was added to the lectin isolate, suggesting that the opsonizing effect was significantly attributable to mannan-binding lectins. The relative contributions of mannan-binding lectins, FN, and transferrin to the opsonizing activity of human serum were examined next. M. tuberculosis H37Rv was preincubated in one of the following: PBS (negative control); 25% human serum (positive control); 25% human serum containing 75 mM mannose, N-acetylglucosamine, or galactosamine; 25% mannosebinding lectin-depleted, FN-depleted, or transferrin-depleted serum; and 25% opsonin-depleted control serum (see Materials and Methods). After preincubation with the various serum preparations, the bacteria were added to CHO cells in serumfree medium. As shown in Fig. 5, the MBP ligands mannose and N-acetylglucosamine inhibited the opsonizing effect of serum by 33% 6 10% and 37% 6 10%, respectively, whereas galactosamine, which is not a ligand for MBP, had no effect. Treatment of the serum with mannan-agarose reduced the opsonization efficacy of the serum by 40% 6 25%; similarly, the opsonizing activity of genetically MBP-deficient serum was 42% 6 8% less than that of the control serum. Treatment of serum with anti-FN-agarose or anti-transferrin-agarose did not significantly reduce its opsonizing capacity (Fig. 5), despite extensive depletion of these opsonins, as revealed by Western blotting (not shown). In contrast to the opsonizing activity of rMBP, FN, and

FIG. 4. Ability of candidate opsonins to enhance the binding of M. tuberculosis H37Rv to CHO cells. (A) Bacteria were preincubated in transferrin (0.5 mg/ml), FN (0.5 mg/ml), rMBP (20 mg/ml), or 20% FBS before incubation with cells in serum-free medium. (B) Similarly, bacteria were preincubated in a crude lectin isolate of FBS (obtained after application of serum to a mannan-agarose column and elution in EDTA), the lectin isolate plus mannan (1 mg/ml), or 20% FBS. All cell-binding data were obtained as described in the legend to Fig. 1. p, P , 0.01 compared to opsonin-free controls; pp, P , 0.05 compared to the opsonin-free control (A) or to lectin isolate-treated bacteria (B).

3900

HOPPE ET AL.

FIG. 5. Contributions of mannan-binding lectins, FN, and transferrin to the opsonizing activity in human serum. M. tuberculosis H37Rv was preincubated in 25% control serum (fresh human serum, serum preincubated with agarose, or serum preincubated with agarose coupled to IgG); in 25% serum containing 75 mM galactosamine, N-acetylglucosamine (NADG), or mannose; in 25% genetically MBP-deficient serum; in 25% lectin-depleted serum; in 25% FN-depleted serum; or in 25% transferrin-depleted serum. Bacteria were then incubated with CHO cells in serum-free medium, and binding data were obtained as described in the legend to Fig. 1; the data were derived from several experiments and are expressed as a percentage of the control serum after subtraction of binding in the absence of serum (relative opsonization). p, P , 0.05 compared to the appropriate control serum.

transferrin, preincubation of M. tuberculosis in 1-mg/ml solutions of BSA, human collagen, low-density lipoprotein, or keratin and chondroitin sulfate had no effect on binding to CHO cells. Binding of M. tuberculosis H37Rv to CHO and endothelial cells was inhibited by MAb 1D1. To identify bacterial determinants involved in binding to mammalian cells, we prepared a panel of MAbs that recognize intact, whole M. tuberculosis and tested the antibodies for the ability to inhibit binding to CHO cells. CHO cells were incubated with M. tuberculosis H37Rv in undiluted hybridoma culture supernatants (DMEM plus 10% FBS) containing the various MAbs. As a negative control, cells were incubated with bacteria in DMEM plus 10% FBS. Of the 11 MAbs tested, most showed little or no inhibition (Table 1). MAbs 1D1 and 3D6, however, inhibited binding by 63% 6 7% and 46% 6 5%, respectively (Fig. 6). Both MAbs also significantly inhibited binding when bacteria were first preincubated in the hybridoma supernatants, washed, and then added to CHO cells in DMEM–10% FBS (results not shown). Binding of bacteria preincubated in 1D1 and 3D6 hybridoma supernatants to porcine endothelial cells was inhibited by 73% 6 8% and 69% 6 9%, respectively. Interestingly, the presence of 1D1 ascites fluid (20 ml/ml) in serum-free medium inhibited the binding of M. tuberculosis H37Rv to CHO cells by 64% 6 7% (Fig. 6A), suggesting that the antigen recognized by MAb 1D1 is involved in opsonic as well as nonopsonic binding of M. tuberculosis H37Rv to CHO cells. Serum-dependent binding of M. smegmatis was similarly inhibited by 65% 6 17% when 1D1 ascites fluid was added, indicating that the 1D1 antigen is also expressed in M. smegmatis and is involved in the binding of these bacteria to CHO cells. In contrast, binding of the Erdman strain to CHO cells was not inhibited by MAb 1D1 (Fig. 6); notably, binding of this strain to CHO cells was also not serum dependent. To identify the antigen recognized by MAbs 1D1 and 3D6, M. tuberculosis H37Rv was solubilized in reducing Laemmli solubilization buffer and the components were resolved on an SDS–15% polyacrylamide gel (Fig. 7A). The M. tuberculosis proteins in lane 1 were stained with Coomassie brilliant blue, whereas those in lane 2 were transblotted onto nitrocellulose and reacted with MAb 1D1. The Western blot shows that the MAb recognized an antigen that migrated as a broad band

INFECT. IMMUN.

FIG. 6. Inhibition of binding of mycobacteria to CHO and endothelial cells by MAb 1D1 and deacylated 1D1 Ag. Mycobacteria were incubated with cells in the absence (A) or presence (B and C) of 10% fresh human serum. MAb 1D1 was added in the form of 1D1 ascitic fluid (20 ml/ml medium); purified 1D1 Ag was deacylated by alkaline hydrolysis and added at 8 mg/ml. (A) CHO cells incubated with bacteria at an MOI of 300:1; (B) CHO cells at an MOI of 100:1; (C) endothelial cells at an MOI of 100:1; (D) CHO cells at an MOI of 100:1 in the presence or absence of 20% FBS. p, P , 0.01, and pp, P , 0.05 compared to antibody- and antigen-free controls. M.tb., M. tuberculosis.

exhibiting an apparent molecular mass of 12 to 16 kDa. No Coomassie blue staining was observed in this region (Fig. 7A, lane 1). The absence of Coomassie blue staining and the diffuse shape of the band suggested that the antigen was not a protein. This inference was supported by the finding that the band did not disappear after M. tuberculosis bacilli were digested with 0.5 mg of proteinase K per ml for 2 h at 37°C prior

FIG. 7. SDS-PAGE and Western blotting of M. tuberculosis H37Rv cell lysates (A and B) and M. smegmatis cell wall fractions (C and D). (A) SDS–15% polyacrylamide gel stained with Coomassie brilliant blue (lane 1) or immunoblotted with MAb 1D1 (lane 2). (B) SDS–10% polyacrylamide gel of lysates treated with proteinase K (lanes 2 and 4) and immunoblotted with MAb 1D1 (lanes 1 and 2) or 4F11 (lanes 3 and 4). (C) Periodic acid-silver-stained SDS– 15% polyacrylamide gel of the 70% ethanol (lane 1) and 2% SDS (lane 2) cell wall extracts and the corresponding HPLC-purified 1D1 Ag (lanes 3 and 4, respectively). (D) Western blot of a similar gel, with lanes as in panel C, reacted with MAb 1D1.

VOL. 65, 1997

FIG. 8. TLC analysis of 1D1 Ag. M. smegmatis cell wall components were applied to silica gel plates developed with chloroform-methanol-water (55:45: 10). (A) Orcinol staining of purified 1D1 Ag (lane 1), 70% ethanol extract (lane 2), and M. bovis BCG PIMs (lane 3). (B) TLC overlay assay with MAb 1D1; lanes 1 to 3 as in panel A. (C) TLC overlay assays reacted with MAb 1D1 (lane 2) or biotinylated rMBP (lanes 3 to 5). Lanes 2 and 3, 70% ethanol extract; lane 4, Man-LAM; lane 5, Ara-LAM. Lane 1 is a control TLC analysis of orcinol-stained 70% ethanol extract.

to electrophoresis on an SDS–10% polyacrylamide gel (Fig. 7B, lanes 1 and 2; note that the antigen migrated as a sharp band at the gel front when a lower-percentage acrylamide gel was used). As a control, the 32-kDa band recognized by MAb 4F11 completely disappeared following the protease treatment (Fig. 7B, lanes 3 and 4). Additional Western blotting experiments with MAb 3D6 yielded identical results (not shown), suggesting that MAbs 1D1 and 3D6 recognized the same antigen. Purification of 1D1 Ag. Due to ease of culture, we used M. smegmatis as the source material. Following the method of Melancon-Kaplan et al. (33) for the fractionation of M. leprae, we sequentially extracted the bacteria with chloroform-methanol, 70% ethanol, and 2% SDS. The lipid extract contained no 1D1 Ag, as assayed by Western blotting and ELISA, but the antigen was present in appreciable amounts in the ethanol and SDS extracts. Figure 7C shows a periodic acid-silver-stained SDS–15% polyacrylamide gel of the ethanol and SDS extracts (lanes 1 and 2, respectively) with numerous components in each, whereas Fig. 7D (lanes 1 and 2) is the corresponding Western blot stained with MAb 1D1. The antigen was subsequently purified from the ethanol extract by reversed-phase HPLC or by silica gel column chromatography, and analysis of pooled, ELISA-positive fractions revealed, by SDS-PAGE and Western blotting, a single, broad band (Fig. 7C and D, lanes 3). This band could also be purified by HPLC from the SDS extract, although at a lower yield (Fig. 7C, lane 4). Chemical and immunological analyses of 1D1 Ag. Purified 1D1 Ag ran as a doublet on TLC after orcinol staining, and it was free of the contaminating glycoconjugates present in the 70% ethanol extract (Fig. 8A, lanes 1 and 2). This doublet was also clearly seen in a TLC overlay assay in which the material was shown to react with MAb 1D1, and the same doublet was present in the unfractionated 70% ethanol extract analyzed in the same manner (Fig. 8B, lanes 1 and 2). Significantly, a band with similar migration on TLC was seen in a sample of authentic PIMs (provided by Patrick J. Brennan, Colorado State University, Fort Collins), although by orcinol staining this appeared to be a complex mixture that contained many slowerand faster-migrating species (Fig. 8A, lane 3). This diversity presumably reflects the presence of polar hexa- and pentamannosides (PIM5 and PIM6), as well as lipophilic dimannosides (PIM2); the latter exhibit further heterogeneity due to a variable number of additional acyl substituents (6). Analysis of these PIMs by TLC overlay with MAb 1D1 revealed a group of five bands, of which two comigrated with the 1D1 Ag doublet (Fig. 8B, lane 3). Significantly, the five immunoreactive bands comprised the slower-migrating, polar PIMs (PIM5 and PIM6),


3901

suggesting that the 1D1 epitope includes at least a dimannoside linked to inositol; in PIM2, the two mannoses are linked separately to the 2 and 6 positions of the inositol ring (6). These results suggested that the 1D1 Ag is a PIM. This conclusion was supported by GC analysis of methanolyzed products of purified 1D1 Ag, which led to the identification of mannose as the only sugar constituent of the glycolipid; in addition, the aqueous phase also contained inositol. The major fatty acid substituents of the antigen were composed of hexadecanoyl (C16:0) and 10-methyl-octadecanoyl (tuberculostearoyl); small amounts of C18:1 and C18:0 were also present in the mixture of fatty acid methyl esters. These data clearly indicated that 1D1 Ag is a member of the ubiquitous family of PIMs, and based on the comparative TLC migrations and immunoreactivities of the antigen and authentic PIMs (Fig. 8A and B), 1D1 Ag is most likely a polar PIM, either PIM5 or PIM6. Binding of 1D1 Ag-coated latex beads to CHO cells. To investigate the role of the 1D1 Ag in cell adhesion, the binding to CHO cells of latex beads coated with the antigen was determined (Fig. 9). Antigen-coated beads adhered strongly to CHO cells in serum-free medium (Fig. 9B), whereas uncoated control beads adhered poorly (Fig. 9A). In the presence of 20% FBS, binding of antigen-coated beads was increased 2.5fold (Fig. 9D and E). When antigen-coated beads were treated with periodate to oxidize carbohydrate residues, serum enhancement of binding was abolished (Fig. 9E), and binding in the absence of serum was also reduced about twofold (Fig. 9E). Binding of uncoated control beads was also improved in the presence of FBS (Fig. 9C), which may be due to nonspecific adsorption of serum opsonins to the beads as a result of incomplete blocking by BSA. The addition of ascites fluid containing MAb 1D1 did not inhibit the binding of periodatetreated beads to CHO cells, likely due to destruction of the epitope recognized by the antibody (in Western blotting experiments the antibody also failed to recognize lysates of periodate-treated M. tuberculosis, further confirming that 1D1 Ag is a carbohydrate-containing molecule). The antibody abolished binding of antigen-coated beads (less than 10 beads per 100 cells) in both the presence and absence of FBS (data not shown). Binding of 1D1 Ag to rMBP. The effect of periodate treatment in eliminating the serum enhancement of binding of the 1D1 Ag-coated beads to CHO cells strongly suggested that the carbohydrate component of the 1D1 Ag mediated the opsonization by serum components, which were likely to include mannose-binding lectins, such as MBP. To confirm this, we analyzed the ability of 1D1 Ag to bind rMBP in a modified ELISA-like binding assay using biotinylated rMBP. We found that rMBP bound to 1D1 Ag in a mannan-inhibitable manner, as it did to whole M. tuberculosis H37Rv and to authentic Man-LAM (the positive control), but not to E. coli LPS (Fig. 10). The extent of binding in this assay to the 1D1 Ag and intact M. tuberculosis was approximately 3.7- and 6-fold less than to Man-LAM. This may indicate that rMBP had a higher affinity for Man-LAM than for 1D1 Ag or, alternatively, that in this assay the rMBP-binding epitopes in 1D1 Ag were not optimally presented. Support for the second possibility derives from a TLC overlay assay using biotinylated rMBP, in which binding of rMBP to 1D1 Ag was strong and equivalent to the binding to Man-LAM, and, significantly, the 1D1 Ag was the only component of the M. smegmatis ethanol extract to which rMBP bound (Fig. 8C, lanes 3 and 4). As expected, in this assay rMBP did not bind to nonmannosylated LAM (Ara-LAM) (Fig. 8C, lane 5).

3902

HOPPE ET AL.

INFECT. IMMUN.

FIG. 9. Binding to CHO cells of latex beads coated with purified 1D1 Ag. Coated and uncoated (incubated in blocking buffer, containing 5% BSA) beads were incubated for 6 h with CHO cells in the absence (A and B) and presence (C and D) of 20% FBS. (A to D) Light micrographs showing uncoated (A and C) and coated (B and D) beads. (E) For quantitation, cells were washed and fixed, and the number of beads bound per 100 cells was determined by light microscopy. Antigen-coated beads were also treated with sodium periodate (0.1 M) at pH 4.5. p, P , 0.0001 compared to uncoated control beads.

Inhibition of binding of M. tuberculosis to CHO cells by deacylated 1D1 Ag. The role of the 1D1 Ag in mediating the binding of intact M. tuberculosis bacilli to CHO cells was analyzed further by incubating the cells with bacteria in the presence of deacylated 1D1 Ag (i.e., PIM without the acyl chains). Deacylated 1D1 Ag was used because it has been shown that the phosphatidylinositol anchor of soluble mycobacterial PIMs inserted into the plasma membranes of mammalian cells (24), which led to a nonspecific inhibition of binding of M. tuberculosis to macrophages (46). In the presence of 20% FBS, the binding of M. tuberculosis to CHO cells was inhibited by 80% 6 12% by deacylated 1D1 Ag, compared to 84% 6 6% by MAb 1D1 (Fig. 6D). Binding in the absence of serum was also inhibited by both deacylated 1D1 Ag and MAb 1D1, by approximately 50%, but this did not achieve statistical significance in this experiment (Fig. 6D). Taken together, these data strongly suggest that the 1D1 Ag

is an adhesin that mediates both direct (nonopsonic) and serum-opsonized binding to CHO cells. The serum opsonin is, at least in part, MBP. Chemically, the antigen is a mycobacterial PIM. This antigen appears to be an adhesin that mediates the binding of M. tuberculosis H37Rv and of M. smegmatis to CHO cells and endothelial cells. DISCUSSION Interactions between pathogen-expressed molecules and specific host proteins can involve direct, opsonin-independent binding of pathogen ligands (adhesins) to cell receptors or indirect, opsonin-mediated binding to cell receptors of appropriate specificity (15). Both modes of interaction have been documented for the binding and invasion of mononuclear phagocytes by M. tuberculosis (40, 41, 45). This distinction may be important in understanding the pathogenesis of tuberculosis because first, the presence and concentration of various opsonins vary widely at different anatomic sites, and second, the mode of entry into cells may determine the fate of the bacterium and its host cell. For instance, it has been shown that M.

VOL. 65, 1997

FIG. 10. Binding of rMBP to 1D1 Ag. Purified 1D1 Ag, Man-LAM, E. coli LPS, and intact M. tuberculosis (M.tb.) H37Rv bacilli were adsorbed onto microtiter plate wells and reacted with biotinylated rMBP, in the presence or absence of mannan (1 mg/ml). Binding was quantitated after incubation with streptavidin-peroxidase. p, P , 0.0001 compared to mannan-inhibited controls and LPS-coated wells.

tuberculosis treated with immune serum (which presumably opsonizes it with specific antibodies) is no longer able to inhibit phagosome-lysosome fusion after invasion of macrophages (1). Similarly, the cytotoxicity of M. tuberculosis for cultured human lung epithelial cells was found to be greatest when invasion occurred in the absence of serum (31). Using M. tuberculosis H37Rv, we found that the ability to infect nonphagocytic cells in vitro was restricted to certain cell types. Of the cells tested, significant association with bacteria was exhibited only by CHO fibroblasts and porcine aortic endothelial cells. It can be speculated that this variability reflects cell-specific expression of suitable receptors, but we have not as yet established the identities of the receptors involved. Notably, both the H37Rv and Erdman strains of M. tuberculosis bound poorly to HeLa cells, contradicting an earlier report which demonstrated efficient invasion of HeLa cells by a variety of M. tuberculosis strains and other mycobacterial species (44). However, in this study it was noted that infection of HeLa cells by mycobacteria was efficient only in the presence of 10 to 40% horse serum and not in human serum; infection rates ranged from .50% of cells in horse serum to ,0.5% in human serum (44). This finding suggests that horse serum contains an unidentified opsonin that enhances the infection of HeLa cells by mycobacteria in vitro. A striking and highly reproducible finding of the binding studies reported here was that infection of both CHO and endothelial cells by strain H37Rv was markedly enhanced by fresh bovine and human serum, suggesting that serum opsonins were able to coat the bacteria and promote binding to the cells. MBP was identified as a candidate serum opsonin, which is of considerable interest in the light of growing reports of the role of this protein as a pattern recognition molecule in first-line, innate host defense (21, 47). By various criteria, including adsorption of serum mannan-binding lectins, addition of mannose or N-acetylglucosamine (both of which are known to bind MBP), and use of human serum from an individual with genetic MBP deficiency, we estimate that MBP contributes approximately 40% of the opsonizing capacity of serum. MBP is a member of the collectin family of soluble C-type lectins, which also includes the lung surfactant proteins SP-A and SP-D, as well as bovine conglutinin and CL-43 (14). All collectins bind sugars that contain equatorial 3- and 4-hydroxy groups, notably D-mannose and N-acetyl-D-glucosamine, which are prominent terminal residues of diverse microbial polysaccharides, and numerous studies have indicated that MBP is an important general opsonin in human serum (reviewed in ref-


3903

erences 14 and 21). Despite evidence for its role in innate immunity, it is notable that heterozygote carriers of variant MBP alleles that result in dominant low MBP concentrations are common in diverse populations throughout the world, which has led to the suggestion that low MBP levels may confer an advantage in certain settings, such as during infection with intracellular pathogens, including mycobacteria, which subvert opsonic mechanisms to invade host cells (19, 21, 30). An association between high serum MBP levels and lepromatous leprosy has been suggested, and sonicates of M. tuberculosis and M. leprae were shown to bind strongly to human MBP (19). These studies and the data presented here support the notion that MBP is an opsonin that can enhance the invasion of host cells by mycobacteria. In addition to MBP, we also found that purified FN and transferrin enhanced the binding of M. tuberculosis to CHO cells. Surprisingly, in view of the role for FN in mediating the binding of M. bovis BCG and M. leprae to nonphagocytic cells in vitro (26, 43), the contribution of FN to the opsonizing capacity of whole serum appeared to be negligible, as revealed by binding studies using FN-depleted serum. A similar result was obtained with serum depleted of transferrin. We conclude that FN and transferrin are not significant serum opsonins in the binding of M. tuberculosis to CHO cells and that FN cannot be considered a universal opsonin for the binding of mycobacteria to nonphagocytic cells. It has been noted that FN-mediated binding to integrin receptors is frequently insufficient to promote bacterial internalization (25). We have characterized a bacterial ligand or adhesin that promotes binding of M. tuberculosis to CHO and endothelial cells. The ligand, denoted 1D1 Ag on the basis of its recognition by MAb 1D1, supported direct as well as serum-enhanced binding to CHO cells, as assessed by binding studies with coated beads. The 1D1 Ag was identified as a polar PIM, which was consistent with our data that rMBP bound to 1D1 Ag in a mannan-inhibitable manner. Identification of the 1D1 Ag as a polar PIM was surprising, in view of the known abundance of LAM in the mycobacterial envelope (8) and the high affinity of LAM for MBP, demonstrated here. LAM is thought to traverse the entire width of the cell wall, with exposure of its terminal sugars at the outer surface of the M. tuberculosis envelope (6, 8, 23), whereas the exact location and degree of surface exposure of the PIMs were, until recently, unclear (8). However, a recent reanalysis of the surface-exposed lipids on the M. tuberculosis cell envelope revealed that PIMs were present in the outermost layers, whereas LAM was notably absent (35, 37). Earlier studies that showed LAM to be present in the outer capsule of M. tuberculosis were based on the use of anti-LAM antibodies (23) that cross-react with neutral, nonacylated arabinomannans (35), which are part of an abundant carbohydrate-protein capsule in which the PIMs are embedded (35–37). Our functional data corroborate these analytical results, indicating that PIMs are sufficiently surface exposed to act as adhesins. It is particularly notable that MAbs 1D1 and 3D6, which were selected for their reactivity to whole bacteria and hence surface epitopes, recognized PIM and not LAM. Indeed, MAb 1D1 appeared to be PIM specific in that it did not cross-react with authentic ManLAM and the binding of this antibody to 1D1 Ag was not inhibited by an excess of free mannose, yeast mannan, or M. tuberculosis mannan (36) in a competitive ELISA (data not shown). These data indicate that a reevaluation of the functional importance of outer capsular PIMs in mycobacterialhost cell interactions is warranted. Interestingly, both the serum enhancement of binding and the role of 1D1 Ag in binding to CHO cells appeared to be

3904

HOPPE ET AL.

strain dependent. Notably, the Erdman strain bound in a serum- and 1D1 Ag-independent manner. The molecular basis for this observed strain variation remains to be determined. In a flow cytometric analysis of exposed surface epitopes, we found, using fluorescent-labeled MAb 1D1, that there were significant differences in surface exposure of PIMs between different M. tuberculosis strains (12). The mode of binding of the H37Rv strain does not, however, appear to be only an oddity of a laboratory strain, since we have identified a clinical isolate (designated GSH-2288, isolated from a urinary catheter; Bacteriology Laboratory, Groote Schuur Hospital) which has the same phenotype (12). A more generalized analysis of the strain dependence of the mode of binding of M. tuberculosis to mammalian cells is clearly warranted and is in progress. The identities of the CHO and endothelial cell receptor(s) for direct or opsonin-mediated binding to PIM are unknown. Direct binding to a mycobacterial PIM implies the presence of a cell surface lectin with specificity similar to that of the macrophage mannose receptor, which is known to be expressed by hepatic endothelial cells (14). Opsonic binding via MBP is presumably to a collectin receptor, such as the SP-A receptor on macrophages and alveolar type II epithelial cells (10). The SP-A receptor binds to the collagen domain of SP-A (10), leaving the carbohydrate recognition domain free to interact with mannose-containing PIMs. Indeed, SP-A enhanced the binding of M. bovis BCG to macrophages, which was blocked by anti-SP-A receptor antibodies (10); presumably a similar interaction could also involve type II epithelial cells (3, 31, 32). The potential in vivo significance of these results is unknown. We have found by transmission electron microscopy of infected CHO cells that many of the bacteria are intracellular (22). However, histologically, M. tuberculosis bacteria are rarely if ever seen in cells other than mononuclear phagocytes in vivo (29). It is possible that invasion of nonphagocytic cells occurs only during brief and specific phases of the disease, such as translocation across the alveolar epithelium at the time of initial exposure (3, 31, 32) and during hematogenous dissemination, or that actual residence time in epithelial or endothelial cells is brief. Our finding that M. tuberculosis binds efficiently to endothelial cells, particularly in the presence of serum opsonins, may reflect the behavior of the pathogen during hematogenous dissemination. In the mouse model, M. tuberculosis bacteria have been found to erode into blood vessel endothelial cells (34), and it has been suggested that CD81 lymphocytes are required for protection because they lyse infected cells that do not express class II antigens, such as endothelial and epithelial cells (9, 34). ACKNOWLEDGMENTS This work was supported by awards under the Glaxo Wellcome Action TB initiative, and by funds from the South African Medical Research Council. We thank R. Alan B. Ezekowitz for useful discussions and for the rMBP and MBP-deficient serum, Patrick J. Brennan for M. tuberculosis LAM and M. bovis BCG PIM, Peter Zilla for the porcine aortic endothelial cells, and Lafras M. Steyn for assistance with mycobacterial cultures. We are indebted to Sylva L. U. Schwager for help with the HPLC purification of 1D1 Ag.

1. 2. 3. 4.

REFERENCES Armstrong, J. A., and P. D. Hart. 1975. Phagosome-lysosome interactions in cultured macrophages infected with virulent tubercle bacilli. J. Exp. Med. 142:1–16. Band, A. H., A. Bhattacharya, and G. P. Talwar. 1986. Lack of Mycobacterium leprae-specific uptake in Schwann cells. Int. J. Lepr. 54:71–78. Bermudez, L. E., and J. Goodman. 1996. Mycobacterium tuberculosis invades and replicates within type II alveolar cells. Infect. Immun. 64:1400–1406. Bermudez, L. E., and L. S. Young. 1994. Factors affecting invasion of HT-29

INFECT. IMMUN.

5. 6. 7. 8. 9. 10. 11.

12. 13.

14. 15. 16. 17. 18. 19. 20.

21.

22. 23. 24.

25. 26. 27. 28.

29. 30.

31. 32.

and HEp-2 epithelial cells by organisms of the Mycobacterium avium complex. Infect. Immun. 62:2021–2026. Bermudez, L. E., K. Shelton, and L. S. Young. 1995. Comparison of the ability of Mycobacterium avium, M. smegmatis and M. tuberculosis to invade and replicate within HEp-2 cells. Tubercle Lung Dis. 76:240–247. Besra, G. S., and D. Chatterjee. 1994. Lipids and carbohydrates of Mycobacterium tuberculosis, p. 285–306. In B. R. Bloom (ed.), Tuberculosis: pathogenesis, protection, and control. ASM Press, Washington, D.C. Bloom, B. R., and C. J. L. Murray. 1992. Tuberculosis: commentary on a reemergent killer. Science 257:1055–1064. Brennan, P. J., and P. Draper. 1994. Ultrastructure of Mycobacterium tuberculosis, p. 271–284. In B. R. Bloom (ed.), Tuberculosis: pathogenesis, prevention, and control. ASM Press, Washington, D.C. Chan, J., and S. H. E. Kaufmann. 1994. Immune mechanisms of protection, p. 389–415. In B. R. Bloom (ed.), Tuberculosis: pathogenesis, protection, and control. ASM Press, Washington, D.C. Chroneos, Z. C., R. Abdolrasulnia, J. A. Whitsett, W. R. Rice, and V. L. Shepherd. 1996. Purification of a cell-surface receptor for surfactant protein A. J. Biol. Chem. 271:16375–16383. Cywes, C., N. L. Godenir, H. C. Hoppe, R. R. Scholle, L. M. Steyn, R. E. Kirsch, and M. R. W. Ehlers. 1996. Nonopsonic binding of Mycobacterium tuberculosis to human complement receptor type 3 expressed in Chinese hamster ovary cells. Infect. Immun. 64:5373–5383. Cywes, C., H. C. Hoppe, M. Daffe´, and M. R. W. Ehlers. Unpublished data. Downing, J. F., R. Pasula, J. R. Wright, H. L. Twigg, III, and W. J. Martin II. 1995. Surfactant protein A promotes attachment of Mycobacterium tuberculosis to alveolar macrophages during infection with human immunodeficiency virus. Proc. Natl. Acad. Sci. USA 92:4848–4852. Drickamer, K., and M. E. Taylor. 1993. Biology of animal lectins. Annu. Rev. Cell Biol. 9:237–264. Falkow, S., R. R. Isberg, and D. A. Portnoy. 1992. The interaction of bacteria with mammalian cells. Annu. Rev. Cell Biol. 8:333–363. Feizi, T., M. S. Stoll, C.-T. Yuen, W. Chai, and A. M. Lawson. 1994. Neoglycolipids: probes of oligosaccharide structure, antigenicity, and function. Methods Enzymol. 230:484–519. Fenton, M. J., and M. W. Vermeulen. 1996. Immunopathology of tuberculosis: Roles of macrophages and monocytes. Infect. Immun. 64:683–690. Fischer, L. J., F. D. Quinn, E. H. White, and C. H. King. 1996. Intracellular growth and cytotoxicity of Mycobacterium haemophilum in a human epithelial cell line (Hec-1-B). Infect. Immun. 64:269–276. Garred, P., M. Harboe, T. Oettinger, C. Koch, and A. Svejgaard. 1994. Dual role of mannan-binding protein in infections: another case of heterosis? Eur. J. Immunogenet. 21:125–131. Hirsch, C. S., J. J. Ellner, D. G. Russell, and E. A. Rich. 1994. Complement receptor-mediated uptake and tumor necrosis factor-a-mediated growth inhibition of Mycobacterium tuberculosis by human alveolar macrophages. J. Immunol. 152:743–753. Holmskov, U., R. Malhotra, R. B. Sim, and J. C. Jensenius. 1994. Collectins: collagenous C-type lectins of the innate immune defense system. Immunol. Today 15:67–74. Hoppe, H. C., C. de Chastellier, and M. R. W. Ehlers. Unpublished data. Hunter, S. W., and P. J. Brennan. 1990. Evidence for the presence of a phosphatidylinositol anchor on the lipoarabinomannan and lipomannan of Mycobacterium tuberculosis. J. Biol. Chem. 265:9272–9279. Ilangumaran, S., S. Arni, M. Poincelet, J.-M. Theler, P. J. Brennan, Nasir-udDin, and D. C. Hoessli. 1995. Integration of mycobacterial lipoarabinomannans into glycosylphosphatidylinositol-rich domains of lymphomonocytic cell plasma membranes. J. Immunol. 155:1334–1342. Isberg, R. R. 1991. Discrimination between intracellular uptake and surface adhesion of bacterial pathogens. Science 252:934–938. Kuroda, K., E. J. Brown, W. B. Telle, D. G. Russell, and T. L. Ratliff. 1993. Characterization of the internalization of Bacillus Calmette-Guerin by human bladder tumor cells. J. Clin. Invest. 91:69–76. Laemmli, U. K. 1970. Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature 227:680–685. Laqueyrerie, A., P. Militzer, F. Romain, K. Eiglmeier, S. Cole, and G. Marchal. 1995. Cloning, sequencing, and expression of the apa gene coding for the Mycobacterium tuberculosis 45/47-kilodalton secreted antigen complex. Infect. Immun. 63:4003–4010. Lucas, S. B. 1989. Mycobacteria and the tissues of man, p. 107–176. In C. Ratledge, J. Stanford, and J. M. Grange (ed.), The biology of the mycobacteria, vol. 3. Academic Press, London, England. Madsen, H. O., P. Garred, J. A. L. Kurtzhals, L. U. Lamm, L. P. Ryder, S. Thiel, and A. Svejgaard. 1994. A new frequent allele is the missing link in the structural polymorphism of the human mannan-binding protein. Immunogenetics 40:37–44. McDonough, K. A., and Y. Kress. 1995. Cytotoxicity for lung epithelial cells is a virulence-associated phenotype of Mycobacterium tuberculosis. Infect. Immun. 63:4802–4811. Mehta, P. K., C. H. King, E. H. White, J. J. Murtagh, Jr., and F. D. Quinn. 1996. Comparison of in vitro models for the study of Mycobacterium tuberculosis invasion and intracellular replication. Infect. Immun. 64:2673–2679.

VOL. 65, 1997 33. Melancon-Kaplan, J., S. W. Hunter, M. McNeil, C. Stewart, R. L. Modlin, T. H. Rea, J. Convit, P. Salgame, V. Mehra, B. R. Bloom, and P. J. Brennan. 1988. Immunological significance of Mycobacterium leprae cell walls. Proc. Natl. Acad. Sci. USA 85:1917–1921. 34. Orme, I. M., and F. M. Collins. 1994. Mouse model of tuberculosis, p. 113–134. In B. R. Bloom (ed.), Tuberculosis: pathogenesis, protection, and control. ASM Press, Washington, D.C. 35. Ortalo-Magne´, A., A. B. Anderson, and M. Daffe´. 1996. The outermost capsular arabinomannans and other mannoconjugates of virulent and avirulent tubercle bacilli. Microbiology 142:927–935. 36. Ortalo-Magne´, A., M.-A. Dupont, A. Lemassu, A. B. Anderson, P. Gounon, and M. Daffe´. 1995. Molecular composition of the outermost capsular material of the tubercle bacillus. Microbiology 141:1609–1620. 37. Ortalo-Magne´, A., A. Lemassu, M.-A. Lane´elle, F. Bardou, G. Silve, P. Gounon, G. Marchal, and M. Daffe´. 1996. Identification of the surfaceexposed lipids on the cell envelopes of Mycobacterium tuberculosis and other mycobacterial species. J. Bacteriol. 178:456–461. 38. O’Shannessy, D. J. 1990. Antibodies biotinylated via sugar moieties. Methods Enzymol. 184:162–166. 39. Ramakrishnan, L., and S. Falkow. 1994. Mycobacterium marinum persists in cultured mammalian cells in a temperature-restricted fashion. Infect. Immun. 62:3222–3229. 40. Schlesinger, L. S. 1993. Macrophage phagocytosis of virulent but not attenuated strains of Mycobacterium tuberculosis is mediated by mannose receptors in addition to complement receptors. J. Immunol. 150:2920–2930. 41. Schlesinger, L. S., C. G. Bellinger-Kawahara, N. R. Payne, and M. A. Horwitz. 1990. Phagocytosis of Mycobacterium tuberculosis is mediated by human

Editor: R. E. McCallum


42. 43.

44. 45.

46. 47. 48. 49. 50.

3905

monocyte complement receptors and complement component C3. J. Immunol. 144:2771–2780. Schnaar, R. L., and L. K. Needham. 1994. Thin-layer chromatography of glycosphingolipids. Methods Enzymol. 230:371–389. Schorey, J. S., Q. Li, D. W. McCourt, M. Bong-Mastek, J. E. Clark-Curtiss, T. L. Ratliff, and E. J. Brown. 1995. A Mycobacterium leprae gene encoding a fibronectin-binding protein is used for efficient invasion of epithelial cells and Schwann cells. Infect. Immun. 63:2652–2657. Shepard, C. C. 1957. Growth characteristics of tubercle bacilli and certain other mycobacteria in HeLa cells. J. Exp. Med. 105:39–55. Stokes, R. W., I. D. Haidl, W. A. Jefferies, and D. P. Speert. 1993. Mycobacteria-macrophage interactions. Macrophage phenotype determines the nonopsonic binding of Mycobacterium tuberculosis to murine macrophages. J. Immunol. 151:7067–7076. Stokes, R. W., and D. P. Speert. 1995. Lipoarabinomannan inhibits nonopsonic binding of Mycobacterium tuberculosis to murine macrophages. J. Immunol. 155:1361–1369. Super, M., and R. A. B. Ezekowitz. 1992. The role of mannose-binding proteins in host defense. Infect. Agents Dis. 1:194–199. Tsai, C.-M., and C. E. Frasch. 1982. A sensitive silver stain for detecting lipopolysaccharides in polyacrylamide gels. Anal. Biochem. 119:115–119. Young, D., R. Lathigra, R. Hendrix, D. Sweetser, and R. A. Young. 1988. Stress proteins are immune targets in leprosy and tuberculosis. Proc. Natl. Acad. Sci. USA 85:4267–4270. Zilla, P., R. Fasol, U. Dudeck, S. Siedler, P. Preiss, T. Fischlein, W. Mu ¨llerGlauser, G. Baitella, D. Sanan, J. Odell, and B. Reichart. 1990. In situ cannulation, microgrid follow-up and low-density plating provide first passage endothelial cell mass cultures for in vitro lining. J. Vasc. Surg. 12:180–189.