Structural features of the arabinan component of the lipoarabinomannan of Mycobacterium tuberculosis.

The recent availability of pure lipoarabinomannan (LAM) from Mycobacterium spp. has resulted in its implication in host-parasite interaction, which events may be mediated by the presence of a phosphatidylinositol unit at the reducing end of LAM. Herein we address the structure of the antigenic, nonreducing end of the molecule. Through the process of 13C NMR analysis of the whole molecule and gas chromatography/mass spectrometry of alditol acetates derived from the differential per-O-alkylated lipopolysaccharide, the majority of the arabinosyl residues were recognized as furanosides. Second, through analysis of per-O-alkylated oligoarabinosyl arabinitol fragments of partially hydrolyzed LAM, it was established that the internal segments of the arabinan component consists of branched 3,5-linked alpha-D-arabinofuranosyl (Araf) units with stretches of linear 5-linked alpha-D-Araf residues attached at both branch positions, whereas the nonreducing terminal segments of LAM consist of either of the two arrangements, beta-D-Araf-(1----2)-alpha-D-Araf-(1----5)- alpha-D-Araf---- or [beta-D-Araf-(1----2)-alpha-D-Araf-(1----]2---- (3 and 5)-alpha-D-Araf----. Since this latter arrangement also characterizes the terminal segments of the peptidoglycan-bound arabinogalactan of Mycobacterium spp., we propose that mycobacteria elaborate unique terminal arabinan motifs in two distinct settings. In the case of the bound arabinogalactan, these motifs provide the nucleus for the esterified mycolic acids, entities which dominate the physicochemical features of mycobacteria and their peculiar pathogenesis. In the case of LAM, these motifs, non-mycolylated, are the dominant B-cell antigens responsible for the majority of the copious antibody response evident in most mycobacterial infections.


Structural Features of the Arabinan Component of the Lipoarabinomannan of Mycobacterium tuberculosis*
The recent availability of pure lipoarabinomannan (LAM) from Mycobacterium spp. has resulted in its implication in host-parasite interaction, which events may be mediated by the presence of a phosphatidylinositol unit at the reducing end of LAM. Herein we address the structure of the antigenic, nonreducing end of the molecule. Through the process of 13C NMR analysis of the whole molecule and gas chromatography/mass spectrometry of alditol acetates derived from the differential per-0-alkylated lipopolysaccharide, the majority of the arabinosyl residues were recognized as furanosides. Second, through analysis of per-0-alkylated oligoarabinosyl arabinitol fragments of partially hydrolyzed LAM, it was established that the internal segments of the arabinan component consists of branched 3,5-linked a-D-arabinofuranosyl (Araf) units with stretches of linear 5-linked a-D-Araf residues attached at both branch positions, whereas the nonreducing terminal segments of LAM consist of either of the two arrangements, &D-Araf-( 1+2)-a-D-Araf-( 1+5)-a-~-Araf+ or [B-~-Araf-( 142)-a-~-Araf-( 1+12+ (3 and B)-a-D-Araf+. Since this latter arrangement also characterizes the terminal segments of the peptidoglycan-bound arabinogalactan of Mycobacterium spp., we propose that mycobacteria elaborate unique terminal arabinan motifs in two distinct settings. In the case of the bound arabinogalactan, these motifs provide the nucleus for the esterified mycolic acids, entities which dominate the physicochemical features of mycobacteria and their peculiar pathogenesis. In the case of LAM, these motifs, non-mycolylated, are the dominant B-cell antigens responsible for the majority of the copious antibody response evident in most mycobacterial infections.
Within Mycobacterium spp. there exists a dominant soluble polysaccharide, termed lipoarabinomannan (LAM),' highly * This work was supported by Grants AI-18357 and 1-27288, and Contract NO1 AI-52582 from the National Institute of Allergy and Infectious Diseases, National Institutes of Health. The NMR conducted at the Colorado State University Regional NMR Center was supported by National Science Foundation Grant CHE 86-16437. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Author of record. To whom correspondence should be addressed.
Tel.: 303-491-6789;Fax: 303-491-1815. ' The abbreviations used are: LAM, lipoarabinomannan; per-0-Me-LAM, per-0-methylated-LAM; Ara, arabinosyl; Gal, galactosyl; Man, mannosyl; Hex, hexose; d, ['HI, deuterium; Me, methyl; f , furanosyl; p , pyranosyl; t, terminal; GC, gas chromatography; GC/ MS, gas chromatography-mass spectrometry; HPLC, high performance liquid chromatography; ald, alditol; EI, electron impact. antigenic, excreted in copious quantities, and with many of the biological attributes long associated with the likes of the 0-antigenic lipopolysaccharides (l), such as suppression of Tlymphocyte activation (2, 3), inhibition of y-interferon activation of macrophages (4), induction of the release of tumor necrosis factor (5, 6), and a generalized inhibition of antigen presentation by antigen presenting cells (7). So pervasive and profound are some of these biological effects that it has been argued, on the one hand, that LAM is the key factor in the intracellular survival of mycobacteria and in much of the pathogenesis of mycobacterioses (1) and, on the other, that it contains endotoxin or endotoxic activity (8), despite the absence of chemical evidence of ketodeoxyoctonate. In fact, Hunter et al. (9) and Hunter and Brennan (lo), following earlier evidence that LAM is phosphorylated (11) and acylated (9), attributed to it many of the physicochemical properties of the 0-antigenic lipopolysaccharides, such as a "membrane anchor" in the form of phosphatidylinositol (9, lo), alkalilabile phosphodiester groups identified as l-phospho-myoinositol (9), a mannan "core," and oligoarabinosyl side chains. The older literature had clearly demonstrated that the arabinan segments of LAM were the basis of its profound B-cell antigenicity in that serological activity was ablated by the action of endoarabinases and mild acid hydrolysis (12,13) and the fact that lipomannan, essentially an arabinose-free version of LAM, was not antigenic (14). In this present report, we address the issue of the structural basis of the antigenicity of LAM, revealing novel oligoarabinofuranosyl arrangements, some of which had been recognized previously as part of the peptidoglycan-bound D-arabino-D-galactan of mycobacteria (15). We speculate that these motifs provide the structural basis of the dominant immunogenicity of LAM, whereas the majority, when appearing as the terminal nonreducing segments of arabinogalactan, provide the templates for esterified mycolic acid, the one entity that dominates the physical properties of mycobacteria, their peculiar pathogenesis and persistence.

EXPERIMENTAL PROCEDURES
Purification of LAM-The isolation of LAM-containing fractions from Mycobacterium tuberculosis H37Ra and primary resolution on columns of DEAE-Sephacel in detergent-containing buffer have been described (9,10). Furthermore to these steps, preparations of LAM, recovered from columns of DEAE-Sephacel and highly pure according to polyacrylamide gel electrophoresis (9), were dialyzed, concentrated on an Amicon flow cell (Amicon 8200; Danvers, MA; 10 kDa molecular mass cut-off membrane), precipitated with 85% ethanol, redissolved in 0.01 M Tris HCl (pH 7.4) containing 0.1% Triton X-100 and applied to a HYDROPORE AX HPLC column (121.4 mm x 25 cm, Rainin, Woburn, MA) equilibrated in the same buffer. The column was eluted with the same buffer followed by a shallow gradient of 0-0.1 M NaC1. Fractions (10 ml) were collected, analyzed for carbohydrate (IO), and positive fractions were re-examined by polyacrylamide gel electrophoresis. Pure lipomannan was eluted with 0.01 M NaCI, followed by the mannosyl pbosphatidylinositols which emerged with 9652 0.02 M NaCl and LAM which was eluted at 0.05 M NaCl. Fractions were pooled, dialyzed, concentrated, and dried. LAM was reprecipitated with 85% ethanol at 0 "C overnight and centrifuged. To remove the last traces of detergent, a solution of pure LAM was passed through a column (2 ml) of Extracti Gel-D (Pierce Chemical Co.) and eluted with H20. Pure LAM thus obtained was the subject of the analyses described below.
Partial Degradation of Per-0-alkylated LAM and Resolution and Analysis of Fragments-The strategy and exact protocols for the random partial depolymerization of per-0-Me-LAM followed by reduction with NaB['HI4 and per-0-deuterioethylation have been described (15). Exact procedures for the separation of the per-0-alkylated oligoarabinosyl arabinitols by HPLC and subsequent GC/MS have also been described (15). Selected HPLC fractions were also separately hydrolyzed, reduced, acetylated, and analyzed by GC/MS (15) in order to confirm substitution arrangements on individual partially-0-alkylated alditol acetates. The per-0-alkylated oligoarabinosyl arabinitols that were the object of 'H NMR analyses were dissolved in hexadeuterioacetone.
NMR analysis on intact LAM was conducted in ['HI20 (15). NMR was performed on a Bruker ACE-300 or -500 NMR at the Colorado State University Department of Chemistry Central Instrument Facility or at the Regional NMR Center. LAM-In the only other contemporary analysis of LAM (ll), it was concluded that substantial amounts of the Ara residues were in the pyranose ring form. In the light of the fact that all Ara residues within arabinogalactan are Araf (15)(16)(17), we wished to re-address the matter. Pure LAM (15 mg) was per-0-methylated as described (16) and an aliquot completely hydrolyzed with 2 M CF,COOH at 120 "C for 1 h, reduced, acetylated, and analyzed by GC/ MS and GC (Table I). Such conventional methylation analysis can establish the ring form of a glycosyl residue only if an 0-methyl substituent appears on C-4 in the case of sugars in the pyranose form or on C-5 for glycofuranoses. All of the mannosyl residues of LAM after methylation contain a methoxyl group at C-4 (Table I) and thus are pyranosyl, an observation consistent with earlier reports (11, 13). In addition, identification of 2,3,5-tri-O-Me-Ara and 3,5-di-O-Me-Ara (Table I) pointed to the presence in LAM of t-Araf and 2-linked Araf, respectively. However, methylation analysis Proof that the product is 2,5-linked Araf rather than 2,4-linked ' Proof that the product is 3,5-linked Araf rather than 3,4-linked amount of 4-linked Arap (see text).

Ring Form of Ara Residues in
Arap is given in Footnote 2.
Arap is given in the text. (Table I) could not help determine the nature of the ring form of the majority of arabinosyl residues, those devoid of LOCH3 groups. In order to address this matter, the native, pure molecule was first subjected to DEPT NMR analysis (Fig.  1). The cluster of signals at 109 ppm can only result from C-1 of a-Araf residues; identical signals in the case of arabinogalactan arise from 5-linked-a-Araf residues (15). The C-1 resonances of 2-linked-a-Araf residues are known to be slightly upfield from the C-1 resonances of 5-linked-a-Araf, and occur at -107 ppm (15). The mannosyl residues are known to be in the a configuration ( l l ) , and, indeed, the three signals at -104, 101, and 97 ppm are attributable to 6linked a-Manp, t-a-Manp, and 2,6-linked a-Manp; the 2,6linked a-Manp signal is upfield, at -97 ppm, due to substitution at C-2; the assignment of the 6-linked-a-Manp and ta-Manp may be reversed. The C-1 of t-P-Araf residues is known to absorb at -102 ppm (15) and, in view of the fact that this signal is noticeably more intense than those at -104 and -97 ppm, the presence of t-P-Araf in LAM was at once suspected. The C-5 resonance of t-P-Araf shown previously to occur at 64.1 ppm (15) was also evident in LAM. Thus, 13C NMR analysis confirmed the furanosyl ring form of the 2linked Ara and the t-Ara as first suggested by methylation analysis. More importantly, 13C NMR demonstrated that the question of 5-linked Araf or 4-linked Arap is settled in favor of 5-linked Araf.
In order to further corroborate the evidence for the dominance of Araf in LAM, the experimental approach successfully applied previously to the related question in the context of arabinogalactan (16,18) was applied. Per-0-methylated LAM was partially hydrolyzed, reduced with NaB['HJ4, 0-ethylated, fully hydrolyzed, again reduced with NaB['HI4, and O-acetylated. The effect of this series of reactions on a hypothetical 4-linked Arap and 5-linked Araf is illustrated in Fig. 2. Thus a hypothetical 4-linked Arap residue cleaved by the partial -CH 8 -CH3 ONLY ANOMERIC CARBONS   I   I  I  I  I  I  I  I  I  I  I  I  I  I  I 2. A scheme for distinguishing 4-linked Arap from 5linked Araf residues. The formation of the four different O-Ac, 0-C,['H],, O-Me alditols, two resulting from pyranosyl residues (2 and Y ) and two resulting from furanosyl residues ( E and A ) , are illustrated. The labels 2, Y , E, and A refer to the labels used for the products/peaks as actually isolated (see Table I1 and Figs. 3 and 4).
hydrolysis step at C-4 (but not at C-1) would be converted In order that both furanosyl and pyranosyl residues be identified, if present, it is necessary that different sets of acid hydrolysis conditions be applied. One set should be appropriate for partial cleavage of the relatively acid-labile furanosyl residues, whereas other chosen conditions should be compatible with the relative acid stability of pyranosides. Thus, referring to Fig. 2, under optimum mild partial acid hydrolysis conditions appropriate for furanosides, neither the 4-linked Arap residues nor R', if R' is a pyranosyl residue, would be cleaved, and, thus, neitherproduct Y nor Z would be produced.
On the other hand, under the stronger partial hydrolysis conditions appropriate for pyranosides, 4-linked Arap and R' each should be partially cleaved, and both products Y and 2 would be generated. Similar reasoning indicated that if R' is furanosyl, product 2 would be formed under both sets of conditions. Likewise, products A and/or E should be formed from 5-linked Araf residues, regardless of the riag form of R, if both conditions of partial acid hydrolysis were applied. The products of all of these possible outcomes are readily distinguishable by MS (18).
Accordingly, two portions of the per-O-Me-LAM were partially hydrolyzed with 2 M CF,COOH (for 1 h) at either 75 or 95 "C. The lower temperature was shown to facilitate cleavage of about one-third of the furanosyl residues, whereas the higher temperature allowed hydrolysis of about one-third of the Manp residues and thus presumably one-third of any Arap residues that may be present. Each sample was reduced with NaB['HI4, pentadeuterioethylated, completely hydrolyzed, reduced with NaB['HI4, acetylated, and analyzed by GC/MS (Figs. 3 and 4). Information leading to identification of the products is provided in Table 11.
At the lower temperature (75 "C), the derivatives expected from 5-linked Araf (A and E, Figs. 2 and 3) were produced in large amounts. Thus, 5-linked Araf as opposed to 4-linked Arap predominates in LAM. Analysis of other products, e.g.
F, G, and J (Table 11), demonstrated the presence of 3,5linked Araf rather than 3,4-linked Arap in LAM. Under the 95 "C partial hydrolysis conditions, few partially alkylated partially acetylated arabinitols were detected; obviously, the majority of the arabinofuranosides were cleaved to monosaccharides at this temperature. However, quantitatively minor amounts of two products, Y (4-0-Ac-1, 5-di-0-Cz[2H]S-2, 3di-O-CH3-arabinitol) and Z (1, 5-di-0-Ac-2, 3-di-0-Cz[2H]s-4-O-Me-arabinitol) ( Fig. 2; Table 11; Fig. 4) were observed, indicative of the presence of 4-linked Arap. Nevertheless, the amounts were extremely low, suggestive of the existence of only a few residues of 4-linked Arap, at most. No derivatives indicative of 3,4-linked Arap were detected by this analysis (Table 11), corroborating results obtained from the application of the milder temperature. Therefore, we conclude that the 2-O-Me-Ara formed during methylation analysis (Table I) Table   11. The mass spectra of products/peaks Z and Y (Table 11) are illustrated. linked Araf and, perhaps, from one or two residues of 4-linked Arap2 Arrangement of Araf Residues in LAM as Derived from Structural Elucidation of Oligoarabinofuranosyl Fragments-The per-0-Me-LAM was hydrolyzed with 2 M CFsCOOH, this time at 75 "C for 45 min, reduced with NaB['H]., and penta-0-deuterioethylated; these partial acid hydrolysis conditions were such that the more acid-stable Manp linkages were not cleaved. The resulting per-0-alkylated oligoarabinosyl-arabinitol mixture was recovered en block from cartridges of Sep-Pak (Waters, Milford, MA) and applied to reverse phase HPLC (15,19). The partially fractionated population of per-0-alkylated oligoarabinosyl-arabinitols was further resolved and analyzed by GC/MS. The application of this series of degradations, derivatizations, separations, and mass spectrometric analyses to a portion of the arabinan segment of LAM is illustrated in Fig. 5.
Conclusions as to the position of the 0-C2['HI5 units are key to the structural interpretations, in that they indicate the original point of attachment of other glycosyl residues. Thus, the O-C2['HI5 groups a t C-2 of glycosyl residue "a" (Fig. 5 0 ) unequivocally established that this residue was originally substituted at C-2 by another glycosyl unit. The presence of ' No derivative of 3-0-Me-Ara (Table I), which could establish its ring form, was observed under either set of acid hydrolysis conditions, since this residue is present in extremely small amounts.

of LAM 9655
When this type of reasoning and other rules governing the interpretation of the EI-MS fragmentation of oligoarabinitol The observation that an internal 2-linked Araf results in both ald J, and ald Jo is borne out through examination of many different oligoarabinosyl arabinitols in both this and the previous study (15). However, the linkage of the internal Ara of compound 12 was confirmed by alditol acetate analysis as documented in Table 111.  23 to the t-Araj and the signal a t 5.07 to the +2-Araf was made possible by recognition that the -2-Araj present in compound 20 was a (Table 11). ald, alditol. alditols (15,19) was applied to a total of 25 such fragments of LAM and the results compared with those derived from a similar analysis of the peptidoglycan-bound arabinogalactan (15), the structures shown in Table I11 emerged.
The Major Arabinan Structural Motifs of LAM-Examination of the structures of these 25 individual oligosaccharide fragments allowed the recognition of four families of oligoarabinosides and thereby four major structural motifs (Fig. 7), three of which, motifs A, B, and D, had been recognized previously as part of the peptidoglycan-bound arabinogalactan (15). Thus, proof of the existence of compounds 5, 6, 10, 11, 18,19, and 21 support the case for motif A; compounds 7, 8, 14, 15, 16, 17, and 22 point to motif B; and compounds 4, 13 and 25 prove the existence of motif D. The presence of structural motif C, which was not encountered previously within arabinogalactan, was deduced in part from the existence of arabinosyl alditol3 which product demonstrated that at least some of the 2-linked Araf residues are glycosidically linked to a linear (nonbranched) 5-linked Araf. A key product, compound 12, confirmed this linkage pattern, i.e. the attachment of a %linked Araf a t C-5 of a 5-linked Araf and, most importantly, indicated that the unit was part of a terminal motif in that a t-Araf was glycosidically linked to C-2 of the 2-linked Araf. The formulation of motif C (Fig. 7) was further aided by the recognition of compound 24 and further confirmed by the structures of compounds 1,20, and 23. Assignment of Anomeric Configurations-Both of the 2linked Araf and 5-linked Araf residues within motif C (Fig. 7) are cy, based on 'H NMR analysis of compound 20 (Table IV).
The t-Ara in motif C is /3 as demonstrated by 'H NMR analysis of compound 12 ( Fig. 6 and Table IV). The +2-Araf residues on both C-3 and C-5 of the branched Araf of structural motif A and the -5-Araf on both C-3 and C-5 of the branched Araf of motif B are in the cy configuration as demonstrated by 'H NMR analysis of compounds 5-8 (Table   IV). The t-Araf residues that are part of in motifs A and C are in the /3 configuration as demonstrated by 'H NMR analysis of compound 1 (Table IV). Finally, the branched 3,5linked-Araf residues in motifs A and B (Fig. 7) are both in the cy configuration according to 'H NMR analysis of compound 9 (Table IV). Thus, assignment of all of the anomeric configurations shown in Fig. 7 was accomplished.
The Application of 13C NMR to the Arabinan Segments of LAM-Previously, we had shown how the application of NMR analysis to solubilized peptidoglycan-bound arabinogalactan allowed assignment of C-1 of the t-Araf units of structural motif A, to 6 101.9 and 6 101.8; C-1 of the 2-linked Araf residue of the same motif was assigned to 6 106.8 and 6 106.6; and C-2 was assigned to 6 88.2 and 6 87.9 (15). The I3C NMR spectrum of LAM was compared in Fig. 8 to that of solubilized arabinogalactan (15) in the range 6 85 to 6 115. Thus, the resonances corresponding to C-1 (6 106.8 and 6 106.6) and C-2 (6 88.2 and 6 87.9) of the 2-linked Araf unit of motif A are readily apparent. In addition, resonances from C-1 (6 106.9) and C-2 (broadening signal at 6 87.9) of the 2linked Araf unit of the closely related structural motif C are also evident. Also, the resonances of the t-Araf units of both motifs A and C were demonstrable between 6 101.7 and 6 101.9; however, they are somewhat obscured by resonances arising from the mannosyl residues. Nevertheless, the presence of motifs A and C in LAM is confirmed by the use of I3C NMR.

DISCUSSION
Some years ago, in reviewing the large body of early elegant work on the structure of the cell wall of mycobacteria, we concluded that the "common antigen of Mycobacterium, Corynebacterium, and Nocardia species" is arabinogalactan "constructed on a main chain of galactopyranan and arabinofuranose (2O)." This summation has proved to be incorrect on two scores. First, Vilkas et al. (21), we (16), and recently Gruber and Gray (17) have established that the arabinogalactan heteropolysaccharide does not contain galactopyranose residues but instead is composed entirely of D-galactofuranosyl and D-arabinofuranosyl units. A significant point of the present work is that the majority of the arabinosyl residues of LAM are also in the furanose form. The evidence, which again is contrary to what prevails in the literature, is based on the fact that NMR of holistic LAM showed abundant signals assignable to Araf and no evidence of Arap. In addition, methylation, followed by partial hydrolysis and subsequent steps, showed the dominance of 5-linked Araf and only minor amounts of 4-linked Arap. Above all, the ring form of the Ara units in the linear di-and trisaccharides were all clearly furanose. The presence of 3,5-linked Araf in M. tuberculosis LAM is unequivocal in light of the data presented in Table I1 and in view of the structures of compounds 6,8,9,11,16,17,and 19 (Table 111) and also considering the rate of acidic cleavage of the 2-0-Me-Ara substituent as compared with other residue^.^ Furthermore, even though we do detect small amounts of 4-linked Arap, the 13C NMR analysis (see peak 109, Fig. 8) as well as the data in Tables I1 and I11 require that 5-linked Araf vastly predominates over 4-linked Arap. In a previous thorough study of the composition of LAM from Mycobacterium smegmatis (ll), the presence of substantial amounts of 4-linked Arap was suggested, and data were presented to the effect that the 2-0-Me-Ara arising from the permethylated product is indicative of 3,4-linked Arap rather than 3,5-linked Araf. Indeed, we5 have confirmed that mild acid hydrolysis of LAM prior to any methylation leads to copious amounts of 1,5-di-O-acety1-2,3,4-tri-O-methyl arabinitol, apparently from t-Arap, and to some 1,3,5-tri-Oacetyl-2,4-di-O-methyl arabinitol, expected from 3,4-linked Arap residues upon cleavage of a substituent at C-4. Thus,  ' brs, Ybroad singlet.
Comuounds 6 and 8 co-eluted from the HPLC. and thus it is not known which anomeric signal results from 6 and whiih results from 8. The assignment of a in both cases in unambiguous.
Compounds 5 and 7 co-eluted from the HPLC, and thus it is not known which anomeric signal results from 5 and which results from 7. The assignment of a in both cases is unambiguous.
Compounds 20 and 13 co-eluted from the HPLC and NMR analysis resulted in a broad singlet at 6 5.00 and 6 5.04 in a ratio of about 3:l. Individual assignments cannot be made, but all of the glycosyl residues must be a-Araf. were established previously (10). Details of the attachment of the arabinan segments to the mannan core are unknown as is the nature of the attachment of the mannan to the phosphatidylinositol anchor. Other groups such as succinyl, lactyl (ll), and inositol phosphate (9,10) are also part of LAM.
the results of Weber and Gray (11) were basically corroborated. On the other hand, similar products were derived from arabinogalactan; and all are agreed (15-17, 21) that Araf rather than Arap predominates in its structure. Likewise, 13C NMR (Fig. l), if not the other evidence, provides unequivocal proof for a preponderance of Araf in LAM and tends to exclude the possibility of hidden "pockets" of Arap. Thus, we must conclude that initial partial acid hydrolysis of LAM leads to unexplained rearrangements of Araf residues resulting in the appearance, as distinct from the reality, of considerable quantities of Arap in LAM. Our present-day image of the body of LAM and of its Araf-containing terminal settings is presented in Fig. 9.
The long-held contention (20) that arabinogalactan is the dominant B-cell antigen of mycobacteria may well be re-PPM evaluated in the light of present work. In a crucial set of studies, Misaki et al. (13) digested the cell walls of several mycobacterial species with alkali, resulting in cleavage of the esterified mycolates and the generation of a solubilized highly purified arabinogalactan. When tested against antiserum raised against whole bacteria, or partially purified cell walls, the arabinogalactan was serologically active. Removal of the majority of arabinofuranosyl through the action of a specific arabinofuranosidase ablated most of the antigen-antibody binding activity (12,13,22), clearly indicating that the multiple branched oligoarabinofuranosyl units represent the responsible determinants or epitopes. Similar evidence from others, combined with the demonstration that soluble, excreted, apparently naturally demycolylated arabinogalactans are antigenic (23), has led to the widely held belief that arabinogalactan is one of the most powerful immunogens of Mycobacterium spp. However, in the light of the evidence that the same terminal branched oligoarabinosyl motifs, the obvious antigen determinants (15), are shared by peptidoglycanbound arabinogalactan and LAM, the role of arabinogalactan in the immunogenicity of mycobacteria must be questioned. It is more likely that LAM, which presents the naked terminal branched pentaarabinosyl unit and its linear arabinosyl variation to the immune system, represents the source of the majority of anti-arabinosyl antibodies in natural and experimental infections. Recently, we have established that the terminal arabinofuranosyl arrangements on the peptidoglycan-bound arabinogalactan are the foci for the attachment of the majority of mycolyl groups (15), an observation in accord with earlier work (24,25) and are, accordingly, expected to be non-immunogenic. On the other hand, LAM, since its generation as a pure product in its native lipopolysaccharide state, has been shown to be highly immunogenic (9) and a powerful antigen in binding to the copious antibodies in cases of human lepromatous leprosy, tuberculosis (10) and bovine tuberculosis (14). Thus, it would appear that mycobacteria, in their biosynthetic economy and exceptional penchant for survival, can produce the same oligoarabinosyl motif for dual purposes, as