Capsular Arabinomannans from Mycobacterium avium with Morphotype-specific Structural Differences but Identical Biological Activity*

The capsules of two colony morphotypes of Mycobacterium avium strain 2151 were investigated, i.e. the virulent smooth-transparent (SmT1) and the nonvirulent smooth-opaque (SmO) types. From both morphotypes we separated a nonacylated arabinomannan (AM) from an acylated polysaccharide fraction by affinity chromatography, of which the AMs were structurally characterized. The AMs from the virulent morphotype, in contrast to that from the nonvirulent form, possessed a larger mannan chain and a shorter arabinan chain. Incubation of murine bone marrow-derived macrophages and human dendritic cells showed that the acylated polysaccharide fractions were potent inducers of tumor necrosis factor-α, interleukin-12, and interleukin-10 compared with nonacylated AMs, which led to only a marginal cytokine release. Further in vitro experiments showed that both the acylated polysaccharide fractions and the nonacylated AMs were able to induce in vitro anti-tumor cytotoxicity of human peripheral blood mononuclear cells. Thus, morphotype-specific structural differences in the capsular AMs of M. avium do not correlate with biological activity; however, their acylation is a prerequisite for effective stimulation of murine macrophages and human dendritic cells.

Numerous bacteria, in particular extraintestinal and invasive pathogens, express capsules that enable them to decorate surface molecules, leading to inhibition of phagocytosis. Other encapsulated intracellular pathogens protect themselves against phagocytic digestion by actively killing macrophages.
Some mycobacterial species contain capsular-like outer layers, which consist mainly of polysaccharides, proteins, and small amounts of lipids (1). First evidence of the presence of a mycobacterial capsule was provided in 1959 by Chapman et al.
(2) as a "capsular space" between the phagosomal membrane of the infected cell and a mycobacterium. Electronmicroscopy led to the term "electron transparent zone." In 1986 the existence of the capsule was confirmed by chemical analyses (3). Unlike other bacterial capsules, the mycobacterial capsule does not seem to offer protection against phagocytosis. In contrast, mycobacteria have developed strategies to penetrate into macrophages, which usually are responsible for initial nonspecific degradation of pathogens but which, during innate immunity, present mycobacteria the optimal environmental conditions for survival and replication (4). Like other capsules, those of mycobacteria consist mainly of polysaccharides. The function of such capsular polysaccharides has not been completely determined up to now. They may protect the bacteria against enzymatic digestion after phagocytosis. Furthermore, capsular polysaccharides possibly trigger the interaction between bacteria and host cells during the first steps of infection, and it has been assumed that they contribute to intracellular persistence and replication. Taken together, capsular polysaccharides are believed to be potent virulence factors.
Three polysaccharides of the capsule of Mycobacterium tuberculosis have been isolated and partly characterized, i.e. a glycogen-like glucan with an estimated molecular mass of 100 kDa, a mannan, and an arabinomannan (AM) 5 (1). The exact localization and arrangement of these polysaccharides remain to be determined. Both glucan and AM were expressed in vivo and in vitro (5,6). It has been shown by immunoelectron microscopy that the AM is localized in the capsule (7). Another modified AM possessing serological activity has been isolated from Mycobacterium smegmatis (8).
The exact structure of the AM (estimated molecular mass, 13 kDa) from M. tuberculosis has not been fully characterized. The proposed structure comprises a mannan backbone substituted by a branched arabinan, which is further modified by Man residues at the nonreducing end. The arabinan is structurally similar to the arabinan of the arabinogalactan located in the cell wall (9,10).
Mycobacterial surface molecules including lipoarabinomannan (LAM) and AMs have been shown to stimulate and activate immune cells. These immunomodulatory properties have ultimately led to the use of such compounds in vaccination against mycobacterial infections and as immunotherapeutic agents in cancer therapy (11)(12)(13)(14)(15)(16)(17)(18)(19). AMs preparations show profound anti-tumor activity in murine experimental cancer models, and mechanistic studies reveal that the anti-tumor activity of AMs depends on distinct immune cell populations and TH1 cytokines (16,17). However, these reported studies were performed on Z-100, an extract from M. tuberculosis strain Aoyama B, which was described as containing AMs as well as LAM. Thus, it remains unclear whether acylation plays a role in the identified biological activities. Thus far, little is known about the structural characteristics and the potential biological role of AMs from other mycobacterial species.
Infection with Mycobacterium avium, a facultative intracellular opportunistic pathogen in humans, is a well studied model for mycobacterial infections. A major phenotypic characteristic of M. avium isolates is their ability to appear in different colony morphotypes, smooth (either transparent or opaque) or rough (reviewed in Ref. 20). Although morphotypes of the same strain are very closely related, previous studies have demonstrated that the smooth-transparent variant is usually better able to survive and grow intracellularly than the opaque morphotype (21,22). With regard to the M. avium morphotypes used in this study (2151 SmT and SmO), this pattern has been confirmed both in murine macrophages and in a mouse model of infection (23,24).
In the current study we purified AMs of SmT and SmO of M. avium strain 2151. To correlate the structural properties with biological activity we used cytokine production by human and murine dendritic cells and induced cytotoxicity of human peripheral blood mononuclear cells (MNCs) as read-out systems. As a result we found distinct structural differences in the AM polysaccharide composition of SmT compared with that of SmO, which did not correspond to differences in biological activity. Our studies also revealed that AMs from M. avium could induce cytotoxicity of human immune effector cells in vitro.

EXPERIMENTAL PROCEDURES
Chemical Reagents-All chemicals were of highest grade. Aqueous solutions for chemical reactions were prepared with Milli-Q water.
Isolation of Polysaccharides-Aqueous phenol (1%, w/v) was added to the freshly harvested wet cells (18 ml g Ϫ1 ), and the suspensions were gently shaken at 4°C for 48 h. After centrifugation at 27,000 ϫ g at 4°C for 20 min, the supernatant was filtrated (0.22 m, Steritop, Millipore), dialyzed against MilliQ-H 2 O (5 ϫ 10 liters) at 4°C, using 3,500 MWCO membrane (SpectraPor, Roth) to remove phenol and low-molecular mass components, and finally lyophilized (yield: 1.5% of the bacterial wet mass). This capsular material was extracted with 10% aqueous Triton X-114 (Sigma, St. Louis, MO) at 37°C as described (26) to eliminate lipoglycans (i.e. phosphatidylinositol mannosides) and lipoproteins. After phase separation, the aqueous layer was precipitated with cold acetone, and the obtained sediment was washed once with cold acetone and then dried under a stream of nitrogen (0.26% of the bacterial wet mass). The aqueous phase was resuspended in a small volume of water and added to 46% (w/v) ammonium sulfate saturation, and incubated at 20 -22°C for 4 h, then at 4°C for 18 h, to precipitate the proteins. The sample was centrifuged (14,000 ϫ g, 4°C, 60 min), yielding a precipitate composed mostly of proteins and a supernatant containing mostly polysaccharides. To remove the salt the sediment and the supernatant were extensively dialyzed at 4°C against aqua bidest. (6 ϫ 10 liters) and freeze-dried. To remove rests of the protein the polysaccharide-rich supernatants were incubated with Proteinase K (10 g ml Ϫ1 ; Boehringer, 50°C, 5 h), dialyzed and finally freeze-dried.
Size Exclusion Chromatography-Proteinase K-treated polysaccharide-rich fractions were applied on a Sephadex G-50 column (3 ϫ 80 cm; Amersham Biosciences) at 20 -22°C and eluted with pyridine/acetic acid/water (4/10/ 1000 by volume) at a flow rate of 0.5 ml min Ϫ1 . The void volume was determined using dextran blue (10 mg ml Ϫ1 ; molecular mass, 2 ϫ 10 6 Da). Separated fractions were detected by a differential refractometer, combined, concentrated on a rotary evaporator, and freeze-dried.
Analyses-For protein quantification a bicinchoninic acid (BCA) microassay was performed according to the supplier's (Pierce) instructions.
The endotoxin contents of fractions A and B isolated by affinity chromatography were established by a chromogenic Limulus amoebocyte lysate assay according to the manufacturer's instructions (Coamatic Chromo-LAL).
Sugar composition and methylation analyses (methylation was repeated twice) were performed as described (29). Separations were performed on a Hewlett-Packard gas chromatograph 438A equipped with a flame ionization detector and a poly-(5% diphenyl-95% dimethyl)-siloxane SPB-5 capillary column (inner diameter, 30 m ϫ 0.32 mm; film thickness, 0.25 m; Supelco). Hydrogen was used as a carrier gas at 60 kilopascals. Alditol acetates were separated at 150°C for 3 min and then, increasing linearly, at 3°C min Ϫ1 to 300°C.
Mass Spectrometry-Identification of partially methylated alditol acetates was performed by GC/MS on a Hewlett-Packard mass spectrometer (HP5989A) equipped with a HP-5MS capillary column. Helium was used as carrier gas at a pressure of 10 p.s.i. The initial temperature of 150°C was increased after 3 min with 5°C min Ϫ1 to a final temperature of 320°C. Electron impact mass spectra were recorded at 70 eV. For chemical ionization, NH 3 was used as the reactant gas. Isolated AMs were analyzed by high resolution electrospray ionization Fourier transform ion cyclotron resonance mass spectrometry (ESI FT-ICR-MS). Measurements were performed on a 7-Tesla-APEX II (Bruker Daltonics) equipped with an Apollo ion source. Spectra were recorded in the positive-ion mode. Samples were dissolved in a mixture of 2-propanol, water, and 30 mM ammonia acetate, pH 4.70 (50:50:0.03, v/v/v), in concentrations of 10 g l Ϫ1 and sprayed with a flow rate of 2 l min Ϫ1 . In experiments using capillary skimmer dissociation the capillary exit voltage was set to 250 V. Theoretical masses of polysaccharide fragments were calculated according to the formula: mass ϭ (a⅐m Ara ) ϩ (b⅐m Man ) ϩ m H 2 O ϩ (c⅐m Na ) Ϫ (c⅐m H ) atomic mass units.
NMR Spectroscopy-NMR experiments were performed on a Bruker DRX AVANCE 600 spectrometer operating at 600.31 MHz for 1 H and 150.96 MHz for 13 C. Samples were repeatedly exchanged with D 2 O (500 l, 99.9% purity, Deutero) with intermediate freeze-drying and then dissolved in 500 l of D 2 O (99.99% purity, Deutero). All one-and two-dimensional spectra were recorded at 27°C. Chemical shifts were expressed in ppm and relative to internal acetone (␦ H 2.225; ␦ C 31.45). COSY, TOCSY, NOESY, and ROESY spectra were measured using data sets (t1 ϫ t2) of 512 ϫ 2048 points. TOCSY and NOESY experiments were carried out in the phase-sensitive mode with mixing times of 100 and 200 ms, respectively. Heteronuclear two-dimensional 1 H, 13 C correlation was measured by HMQC and HMBC experiments with data sets of 256 ϫ 2048 points. For homo-and heteronuclear correlations experiments, 128 scans were acquired.
Isolation and Stimulation of Murine Bone Marrow Macrophages-Bone marrow-derived macrophages were isolated from 8 -10-week-old C57BL/6 mice (Charles River) as described (30); 5 ϫ10 5 macrophages (10 6 ml Ϫ1 ) were stimulated with the isolated polysaccharide fractions A and B in concentrations of 1 and 10 g ml Ϫ1 at 37°C, 5% CO 2 for 24 h. Supernatants from stimulated murine macrophages were examined for the presence of TNF-␣. Detection was performed with an anti-mouse TNF-␣ kit according to the manufacturer's instructions (DuoSet mouse TNF-␣/TNFSF1A, R&D Systems).
Generation of Human Monocyte-derived DC-Peripheral blood mononuclear cells were isolated from citrate blood of healthy donors by discontinuous gradient centrifugation (Biocoll (separating solution), Biochrom). Monocytes were elutriated by counterflow centrifugation and concentrated to 10 6 cells ml Ϫ1 RPMI 1640 medium containing 5% fetal calf serum (Linaris), 100 units ml Ϫ1 penicillin, and 100 g ml Ϫ1 streptomycin. Differentiation into immature DC was achieved by a 7-day culture in the presence of granulocyte/macrophage-stimulating factor (500 units ml Ϫ1 ) and IL-4 (500 units ml Ϫ1 ). 10 6 DC ml Ϫ1 were stimulated with 10 g ml Ϫ1 isolated polysaccharide fractions A and B at 37°C, 5% CO 2 for 18 h. The supernatants were examined for the presence of TNF-␣, IL-10, and IL-12 (Ready-Set-Go, human IL-12 ELISA, eBiosciences) by using enzyme-linked immunosorbent assay kits according to the manufacturer's instructions.
Stimulation of Human MNC-Human MNC were stimulated with polysaccharide fractions A and B in concentrations of 10 and 40 g ml Ϫ1 , Mycobacterium bovis BCG (Connaught substrain, Immucyst) with an multiplicity of infection of 0.04 cell Ϫ1 and phosphate-buffered saline (100 l) were added. Cells were cultured for 7 days in 6-well microtiter plates at 37°C, 5% CO 2 .
Chromium Release Assay-Cytotoxicity of stimulated human MNC against tumor cells was determined in a standard chromium release assay. Target cells were harvested, washed, and labeled with 50 Ci of Na 2 51 CrO 4 (Hartmann)/10 6 cells for 1.5 h at 37°C. After washing, the cells were resuspended with a final concentration of 5 ϫ 10 4 cells ml Ϫ1 . Effector cells were harvested and added to a total of 100 l of target cells at effector/target ratios of 40:1, 20:1, and 10:1.
The radioactive content of the supernatant was measured in a ␥-counter (Berthold) after incubation for 5 h at 37°C, 5% CO 2 . The specific lysis was determined according to the formula: specific lysis (%) ϭ (R eff Ϫ R spo )/(R max Ϫ R spo ) ϫ 100. By adding 100 l of medium, the spontaneous release of radioactivity (R spo ) was defined. The maximum release (R max ) was achieved by adding 100 l of lysis solution. The experimental release of radioactivity, which was induced by the effector cells (R eff ), was measured in triplicates.

RESULTS
Extraction and Purification of AM-Applying a mild aqueous phenol extraction method (31), capsular material was extracted from the envelope of M. avium. SDS-PAGE of this phenol extract showed numerous proteins after Coomassie staining. Successive blotting and staining with ConA-conjugated peroxidase revealed the occurrence of LAM, lipomannan, and phosphoinositolmannan (not shown). The lipoglycans not only are membrane-associated (32) but also are thought to be present in the outer surface of the bacilli (33) and thus extractable under mild conditions. Routine monosaccharide analysis showed the presence of sugars in the phenol extract, namely Ara, Man, Glc, Gal, and two different O-methylhexoses. To separate the lipoglycans, an additional extraction utilizing the detergent Triton X-114 was performed (33). The resulting detergent phase included LAM, lipomannan, and phosphoinositolmannan, and the aqueous phase comprised proteins and carbohydrates. Precipitation with ammonium sulfate separated proteins from carbohydrates. The protein-containing sediments are presently being investigated separately. The supernatants (containing carbohydrates) were further fractionated by size exclusion chromatography on Sephadex G-50. Three base-line-separated fractions were obtained, the last of which was identified as salt. Fraction 1 (eluting in the void volume) contained Glc and Gal in an approximate molar ratio of 1:3. Fraction 2 comprised Ara and Man in an approximate molar ratio of 5:4 (preparation of M. avium 2151, SmO) and 10:7 (preparation of SmT), as well as Glc and O-methylhexose in minor amounts. Fractions 2 of both SmO and SmT were separated on ConA-Sepharose, from which ConA-nonreactive acylated polysaccharides (fraction A) and ConA-reactive arabinomannans (fraction B) were isolated (Fig. 1).
Fractions B consisted exclusively of Ara and Man and thus were pure AMs. A distinct difference in the Ara/Man ratio was identified between the AMs obtained from SmO and from SmT. In the AMs of SmO, an approximate molar ratio of Ara/ Man of about 1:1 was identified (2070 nmol mg Ϫ1 Ara; 2201 nmol mg Ϫ1 Man), whereas in the AMs of SmT a ratio of ϳ3:10 (692 nmol mg Ϫ1 Ara; 2319 nmol mg Ϫ1 Man) was found. Methylation analyses of both AMs revealed the presence of the same substituted sugars but in different ratios (data can be found in the supplemental information, Table S1). In both AMs four differently substituted Araf and six differently substituted Manp residues were identified. However, there was a clear structural difference between these polysaccharides. In agreement with the results of neutral sugar analyses, a total of 60% hexose residues was identified in the AMs from SmT and only 41% in the AMs from SmO. With regard to the pentose residues, the polysaccharides differed in the content of both the 5and 2-substituted Araf residues. The AMs from SmT contained ϳ18% 5-substituted Araf and the AMs from SmO 28%. The quantity of 2-substituted Araf was less in SmT (ϳ7%) than in SmO (ϳ12%). Regarding hexose residues there was a clear difference in the amount of 6-substituted Manp. The content of this residue in the AMs from SmT (19%) exceeded that in the AMs from SmO (6%). Branched 2,6-disubstituted Manp residues were twice as abundant in the AMs from SmT (ϳ12%) than those from SmO (6 -7%). 2-Substituted Manp (ϳ6% in SmT and SmO) as well as 3-substituted Manp (5% in SmT, ϳ4% in SmO) occurred in both AMs in similar content. t-Manp residues were also present in similar concentrations (SmT, 15%; SmO, 17%). Thus, the results of methylation analyses were consistent with structural differences between both AMs as indi- cated by compositional analyses. The polysaccharide from SmT possessed a larger mannan (ϳ60%) and shorter arabinan (ϳ40%), and the AMs from SmO comprised a truncated mannan (41%) and a larger arabinan (ϳ59%).
NMR Spectroscopy-NMR experiments identified a pyranosidic ring form for all hexoses and a furanosidic configuration for all pentose/arabinose residues. Signals at 0.9 and 1.3 ppm demonstrated the presence of fatty acids in fraction A, showing that at least some molecules were acylated. Thus, chemical and NMR spectroscopy analyses revealed that fractions A contained lipoglycans, probably including LAM. These structures were not investigated further.
One-dimensional 1 H-NMR spectra of the AMs from SmT and SmO (Fig. 2) showed at least 10 partially overlapping anomeric signals (A-J, 5.5-4.9 ppm) with chemical shifts typical for the ␣-configuration. The presence of fatty acids could be excluded because of missing signals around 0.9 ppm (-CH 3 ) and 1.3 ppm (-CH 2 -). In both spectra very similar chemical shifts, but with different intensities, were identified. Two-dimensional NMR experiments, namely COSY and HMQC (supporting information in supplemental Figs. S1 and S2, respectively), allowed overlapping signals to be distinguished. In the case of the AMs of SmT, the resonance at 5.13 ppm (Fig. 2a) comprised three anomeric signals. One of these (at 5.123 ppm) correlated with an H-2 resonance at 4.300 ppm (G1,2 in supplemental Fig.  S1a), another one at 5.123 ppm with an H-2 resonance at 4.042 ppm (F-c1,2), and a third at 5.132 ppm was correlated to an H-2 resonance at 4.034 ppm (F-b1,2). Furthermore, two overlap-ping signals around 5.09 ppm were distinguished with chemical shifts at 5.089 and 5.098 ppm (H-b1,2 and H-a1,2, respectively). Also, the most abundant resonance at 5.045 ppm (I in Fig. 2) comprised two signals at 5.031 and 5.048 ppm (I-b1,2 and I-a1,2, respectively, in supplemental Fig. S1a).
Likewise, overlapping signals in the 1 H-NMR spectrum of the AMs from SmO could be differentiated. The signal at 5.12 ppm (Fig. 2b) comprised two resonances at 5.116 and 5.132 ppm (G1,2 and F1,2, respectively, in supplemental Fig. S1b). Furthermore, the signals H and C (at 5.09 and 5.17 ppm, respectively) could be differentiated between H-a, H-b, and H-c and between C-a and C-b.
Proton chemical shifts were assigned by COSY and TOCSY spectra and carbon chemical shifts by HMQC and HMBC spectra. In both AMs, all sugar residues possessed the ␣-configuration. For all Man residues, the ␣-configuration was assigned in HMQC experiments without decoupling (spectra not shown) by determination of the coupling constants (J H1,C1 Ͼ 170 Hz). The identification of ␣-configured Ara residues was possible because of the chemical shifts of the respective C-1 atoms. In the AMs from SmT, six different Ara residues were identified (A, B, C, G, H-a, and H-b, in supplemental Table S2), all of which were furanoses. Chemical shifts of single 13 C atoms were compared with those of unsubstituted monosaccharides (35) and resulted in the identification of substituted carbon atoms. The C-2 resonances of Ara residues A and B were shifted downfield (ϳ⌬ 2.4 ppm), thus proving that these residues were substituted at position 2. In residue G the C-3 resonance was shifted downfield (ϳ⌬ 2.4 ppm), identifying this sugar as 3-substituted Ara. Whether this residue was a 3,5-disubstituted Ara, which was detected in methylation analysis (ϳ16 mol %, supplemental Table S1), could not be clarified. The C-5 resonances of Ara residues H-a and H-b were shifted downfield (ϳ⌬ 2.2 ppm) and therefore identified a substitution at O-5. Eight different Man residues were identified (D, E, F-a, F-b, F-c, I-a, I-b, and J, supplemental Table S2), all of which were pyranoses. Chemical shifts of the C-2 of Man residues D and E proved their substitutions at O-2 (⌬ 3.6 ppm for C-2 of D and ⌬ 2.9 ppm for C-2 of E) (36). According to methylation analysis, which identified 2,6-disubstituted in addition to 2-substituted Man residues, for Man D and Man E an additional substitution at O-6 could be possible. However, because of strong overlapping, such substitution could not be confirmed by NMR spectroscopy. On the contrary, Man F-a, Man F-b, and Man F-c were clearly identified as 2,6-linked Man residues because of downfield shifts of each C-2 (⌬ 8.6 -8.7 ppm) and C-6 (⌬ 4.6 -5.6 ppm). Resonances of C-2, C-3, C-4, and C-6 of Man I-a and Man I-b were not shifted at all (35); therefore, these were terminal residues. C-6 of Man J was shifted downfield (⌬ 1.3 ppm) identifying it as 6-substituted Man.
In the same way substitutions were assigned for sugar residues of the AMs from SmO (supplemental Table S3). The sugar sequences of the AMs were identified by ROESY experiments. In the spectrum of the AMs from SmT (supplemental Fig. S3a) both protons D1 and E1 showed NOE connectivities to proton B2. This demonstrated 2-substituted Manp residues being linked to position 2 of Araf residues. Furthermore, interresidual effects between proton I-a1 and proton F-c2 were observed showing t-Manp linked to position 2 of 2,6-disubstituted Manp. An NOE contact between proton F-a1 and proton J6 demonstrated the substitution of 6-substituted Manp with 2,6-Manp. Additionally, an interresidual contact between proton J1 and proton F6 gave evidence for a 1 3 6 chain built up of 6-substituted and 2,6-disubstituted Manp, the last of which is substituted at position O-2 with t-Manp. The linkage of t-Manp to position O-2 of 2-substituted Manp was identified by an NOE contact between proton I-b1 and proton D2.
In the ROESY spectrum of the AMs from SmO (supplemental Fig. S3b), interresidual contacts were observed between proton B1 and proton G3, demonstrating the linkage of 2-substituted Araf to position 3 of 3-substituted Araf. The NOE connectivity between proton D1 to proton B2 proved the 1 3 2 linkage between 2-substituted Araf and 2-substituted Manp. An interresidual contact between proton G1 and proton C5 demonstrates that 5-substituted Araf was linked to position 5 to 3-substituted Araf. NOEs between proton J1 and E6 and between I-b1 and F2 show 6-Manp linked at position 6 and t-Manp linked at position 2 of 2,6-Manp.
Mass Spectrometry-To determine the molecular masses of the AMs, mass spectrometry was performed. The spectra comprised a complex of molecules each in different charge state. To simplify the interpretation, the mass spectra were charge-deconvoluted leading to the presence of only neutral species. Fig.  3 depicts the deconvoluted ESI mass spectrum of the AMs from SmO. Numerous signals reflected the heterogeneity of the polysaccharide and the fragmentation induced by capillary skimmer dissociation. The spectrum comprised a broad distribution of ions between 700 and 4500 atomic mass units, with characteristic mass differences of ⌬ m ϭ 22, 30, 132, and 162 atomic mass units. Mass differences of ⌬ m ϭ 22 atomic mass units were due to additional sodium adducts. Mass differences of ⌬ m ϭ 162 atomic mass units corresponded to the mass of one hexose residue. Appropriately, differences of ⌬ m ϭ 132 atomic mass units were due to pentose residues. Based on the results of neutral sugar analyses, hexose was Man and pentose was Ara. ⌬ m ϭ 30 atomic mass units resulted from the "exchange" of one Ara for one Man residue. There were no further mass differences, consistent with the presence of different monosaccharides or contaminations. Each signal represented an ion of the polysaccharide, comprising Ara and Man residues only. Based on the neutral sugar analyses, mass peaks could be explained by different numbers of Ara and Man residues. Combinations were calculated and compared with the measured mass peaks, and measured and calculated masses were in excellent agreement (supplemental Table S4). It was obvious that all abundant signals represented fragments comprising more Ara than Man. In each fragment Ara was at least twice as abundant as Man. The most abundant peak was detected at 2558.791 atomic mass units representing the mass of a molecule comprising 14 Ara and 4 Man residues (ϩ2 sodium ions). The expanded region (Fig. 3) shows completely separated isotope distributions including lower abundant signals, which can also be explained as sodium adducts consisting of 15 Ara and 3 Man (2528.785 atomic mass units) as well as 13 Ara and 5 Man (2558.804 atomic mass units) residues.
The deconvoluted ESI mass spectrum of the AMs from SmT (supplemental Fig. S4 and Table S5) also showed numerous signals between 1100 and 4600 atomic mass units, with characteristic differences of 22, 30, 132, and 162 atomic mass units, which were caused by masses of sodium, Ara, and Man, respectively. One single fragment consisted exclusively of Man (fragment 2 in supplemental Table S5). Such a fragment was not detected in the AMs from SmO. Furthermore, in comparison with the AMs from SmO, this spectrum showed fragments comprising more Man than Ara residues (fragments 2, and 8 -11 in supplemental Table S5); however, also one cluster of fragments with more Ara residues was present (fragments 1 and 3-6). This was in good agreement with the results of neutral sugar and methylation analyses (supplemental Table S1).
ESI MS analysis gave new information about structural features of the AM. The most abundant fragments of both AMs consisted of Ara and Man. There are hardly any fragments comprising only Ara or only Man. Few signals (fragments 3, 7, and 15 in supplemental Table S4 and 1, 3, and 6 in supplemental Table S5) were present in both AMs representing fragments with Ara residues dominating. The AMs from SmT showed fragments with a maximum of 20 Man residues (fragment 11, supplemental Table S5), whereas in the spectrum of the AMs from SmO only fragments with not more than seven Man residues were found (fragments 15 and 16, supplemental Table  S4). Compared with the AMs from SmO, the number of Man residues was significantly higher in the AMs from SmT, which was consistent with the data from chemical analyses. Based on these data, structural models of the AMs of M. avium 2151 SmO and SmT are proposed (Fig. 4).
Induction of Cytokines-To define the biological activity of the M. avium-derived AMs and acylated polysaccharides, we used in vitro stimulation assays of murine and human immune cells. Stimulation of murine bone marrow-derived macrophages as well as human DC with lipoglycan fractions (fraction A in Fig. 1) and AMs (fraction B in Fig. 1) from SmO and SmT resulted in the production of both immunostimulatory cytokines (TNF-␣/IL-12) and the immunoregulatory cytokine IL-10 ( Fig. 5; other data can be found in supplemental Fig. S5). TNF-␣ and IL-10 levels were comparable with those induced by LPS used as positive control. In contrast, AMs only slightly induced cytokine release even after stimulation with a high dose of 10 g/ml. All fractions were tested for endotoxin contaminations via a LAL (Limulus amebocyte lysate) assay. Lipoglycan fractions contained LPS in concentrations of 100 pg g Ϫ1 sample (SmT) and 81 pg g Ϫ1 (SmO). In AMs only marginal endotoxin concentrations of 15 pg g Ϫ1 sample (SmT) and 3 pg g Ϫ1 (SmO) were detected.
In summary, the cytokine measurements revealed that the induction of proinflammatory TNF-␣ as well as IL-12 and antiinflammatory IL-10 depends on structural features, namely acylation of the stimulus. This was shown in murine macrophages as well as in human dendritic cells.
Anti-tumor Cytotoxicity-Because of their immunomodulatory properties, mycobacterial AMs are used in vaccination and immunologic cancer therapy. Therefore, to further assess the biological activity of the lipoglycan fractions (fraction A in Fig.  1), and AMs (fraction B in Fig. 1), human MNC were stimulated and tested for cytotoxicity against the human urinary bladder carcinoma cell line T24. After 5 h of cocultivation of effector and 51 Cr-labeled target cells, the released radioactivity was measured, and specific lysis was calculated. As depicted in Fig. 6 both AMs and lipoglycan induced substantial anti-tumor cytotoxicity of human MNC. All tested fractions reached between 40 and 60% of the activity achieved with BCG, which has been described as a potent stimulator. No significant difference between AMs and lipoglycan or between SmO and SmT was observed. Interestingly, this indicates that activation of human cytotoxic effector cells, unlike cytokine release of dendritic cells, is independent of the lipid domain.
Repeated experiments revealed that stimulation with AMs and lipoglycan fractions does not induce apoptosis of monocytes (data not shown). An apoptosis-inducing effect can be excluded for concentrations of 10 g ml Ϫ1 stimulus.

DISCUSSION
The mycobacterial cell envelope represents a complex and unique structure compared with cell envelopes of other prokaryotes, as it is composed mainly of polysaccharides, lipoglycans, proteins, and peptidoglycan. In addition, the cell envelope of some mycobacterial species is surrounded by an outermost layer described as a capsule. Constituents of this capsule represent potential virulence factors because they could be involved in the first steps of host cell infection or in cell protection of intraphagosomal mycobacteria. M. avium, a facultative intracellular opportunistic pathogen in humans, is a well studied model for mycobacterial infections. A major phenotypic characteristic of M. avium isolates is their ability to appear in differ-  were cultivated with the stimuli (10 g ml Ϫ1 ) for 7 days at 37°C in 5%CO 2 . 51 Cr release of labeled tumor cells was measured after 5 h of cocultivation with stimulated effector cells. Specific lysis was determined according to the formula "specific lysis (%) ϭ (R eff Ϫ R spo )/(R max Ϫ R spo ) ϫ 100" where R eff is experimental release, R spo spontaneous release, and R max maximum release. Effector/target ratios were 10:1, 20:1, and 40:1. Ctrl., control (unstimulated cells); BAK, BCGactivated killer cells. ent colony morphotypes: smooth (either transparent or opaque) or rough (reviewed in Ref. 20). Although morphotypes of the same strain are very closely related, previous studies have demonstrated that the smooth-transparent variant is usually better able to survive and grow intracellularly than the opaque morphotype (38).
To elucidate whether structural differences in the mycobacterial capsule composition may explain the different biological activities of these isolates, the structural characteristics of capsular polysaccharides from M. avium strain 2151 SmT and SmO were determined. Neutral AMs were separated from a lipoglycan fraction, most likely comprising LAM. It is noteworthy that this fraction was recovered after a mild extraction technique using 1% phenol. Usually for extraction of lipoglycans the French press method is used. Our finding supports the hypothesis that LAM is not only membrane-associated (37) but also is located in the outermost layer of the mycobacterial cell wall as first proposed by Rastogi (38).
Analysis of the AMs revealed structural differences between SmT and SmO. The AMs from SmT possessed a relative molar ratio of Ara:Man of 1:4, whereas that from SmO contained Ara and Man in similar proportions. Substitution patterns of Ara and Man were in accordance with published structures of other mycobacterial AMs (39). Interestingly, the AMs isolated herein showed lower molecular masses compared with that described for M. tuberculosis. A further structural characterization was achieved by various NMR experiments; however, the occurrence of a number of overlapping signals did not allow a full assignment of all protons. In both AMs, the NMR data proved a (1 3 6)-linked mannan main chain with alternating 6-substituted and 2,6-disubstituted Man residues, the latter being substituted at position 2 by a terminal Manp. A (1 3 5)-linked arabinan backbone was substituted in position 3 with a (1 3 5)-linked side chains. It is not clear whether 5-substituted and 3,5-disubstituted Araf residues were alternating or whether (1 3 5)-linked chains were located between the branching points. The length and position of the side chains could not be determined. t-Manp residues represented "cap" structures of the side chains, directly substituting (1 3 5)-linked Araf and branching from (1 3 2)-linked Manp. Probably, also cap structures composed of (1 3 3)-linked Manp residues were present. The linkage of the arabinan to the mannan domain remained unidentified.
Mass spectrometric analyses of the isolated AMs gave further information about their structural features. Based on chemical and NMR data, particular fragments in ESI FT-ICR mass spectra could be assigned as side chains. Fragments composed exclusively of Manp residues were identified only in the case of the AMs from SmT. A reason for this could be that this mannan showed a different fragmentation behavior due to partial structures lacking branches, compared with the arabinan possessing larger side chains, i.e. such long-chain unbranched mannan fragments to a lower degree. For the arabinan possessing side chains, a stronger fragmentation was expected. Therefore, most of the identified fragments could be assigned as arabinan side chains (with more or less portions of the main chain). However, not every fragment represented a side chain. It had to be considered that each side chain of the postulated structure pos-sesses an even number of mannose residues. The presence of signals corresponding to fragments with an odd number of mannose units could be explained by the loss of Man residues.
The AMs from SmO comprised a truncated mannan and a larger arabinan compared with the AMs from SmT. Because of the larger arabinan, a number of fragments could be explained in part as arabinan side chains (with or without units of the main chain). Further structural differences between the AMs from SmO and SmT could not be identified but were not excluded.
Aspects of the biological activities of all isolated AMs and lipoglycans were examined to determine the relevance of the lipid versus the sugar moieties and the possible differences between components of virulent SmT and nonvirulent SmO.
Lipoglycan fractions induced substantial amounts of proinflammatory TNF-␣ and immunoregulatory IL-10, whereas lipid-free AMs only marginally induced TNF-␣ production, which was not due to LPS contamination. Measurements of IL-10 production by murine bone marrow-derived macrophages after stimulation with lipoglycan fractions and AMs showed comparable results. After stimulation with the lipoglycan fractions, high concentrations of IL-10 were measured.
Because IL-12 is a key cytokine inducing strong TH1 immune responses and anti-tumor immunity, the capacity of our isolated components to induce production of bioactive IL-12 p70 in human DC was determined (40). Although production of bioactive IL-12 often requires co-stimulation with cytokine IFN-␥, for example, lipoglycans but not AMs induced secretion of substantial amounts of IL-12 p70 in human DC.
The observation that a similar biologic activity was observed when human and murine macrophages were stimulated with acylated AMs isolated from both strains indicates that these structures are biologic response modifiers expressed by both morphotypes. However this also demonstrates that acylated AM are not responsible for the major differences with respect to macrophage activation observed in infection experiments with viable bacteria. The highly replicative smooth-transparent morphotype of M. avium strain 2151 induced significantly less phosphorylation of the mitogen-activated protein kinases p38 and ERK1/2 than the smooth-opaque morphotype of the same strain, which was gradually eliminated from macrophage cultures. These differences have also been seen on the level of cytokine formation (41,42). To date, the molecular basis for these differential effects are not understood: A recent report by Bhatnagar and Schorey (43) indicates that the lack of glycopeptidolipids on the bacterial surface may correlate with an increased macrophage activation due to the presence of other macrophage-activating cell wall components.
Mycobacterial AMs have been described as potent biological response modifiers. For example, Z-100 (an AM-rich extract from M. tuberculosis) showed anti-metastatic activity in a murine B16 melanoma model (14). However, the preparations used in those studies also contained LAM, and thus the active compound remained unclear. Our studies found similar immunomodulatory activities of highly purified AMs from M. avium. We stimulated human MNC with both lipoglycan fractions and AM, and tested cells for activity against the human urinary bladder carcinoma cell line T24. Both lipoglycans and AMs induced profound anti-tumor cytotoxicity in the range of 40 to 60% of that obtained by M. bovis BCG. BCG has been described as inducing strong natural killer cell-mediated tumor cell lysis in this system and serving as a positive control (44,45).
In conclusion, our experiments showed that both lipoglycan fractions and AMs have biological response modifying activity. However, a morphotype-specific difference was not detectable. Experiments with murine macrophages and human dendritic cells illustrated the importance of acylation for activation of myeloid immune cells. Cytokine induction by acylated mycobacterial AMs have been described (46 -48). Gilleron et al. (49) showed that the lipid anchor in AraLAM from M. smegmatis is of key importance regarding cytokine induction. Other work supports the hypothesis that biological effects depend on the degree and chemical structure of capping motifs and particular substitutions on the arabinan moiety, i.e. succinate (50). Our investigations showed that neither in AMs nor in lipoglycan fractions obtained from M. avium morphotypes do such substitutions occur. Our data confirm that lipid moieties such as those present in LPS or bacterial lipoproteins are essential for cytokine induction.