Location of the Mycolyl Ester Substituents in the Cell Walls of Mycobacteria*

The question of the precise location of mycolic acids, the single most distinctive cell wall entity of members of the Mycobacterium genus, has now been addressed. The free hydroxyl functions of the arabinogalactan component of the mycobacterial cell wall were O-meth- ylated under conditions in which the mycolyl esters were not cleaved. Subsequent replacement of the my- colyl functions with 0-ethyl groups resulted in an acid-and base-stable differentially 0-alkylated surrogate polysaccharide, more amenable to analysis. Complete hydrolysis, reduction, acetylation, and gas chromatog- raphylmass spectrometry revealed the unexpected finding that the mycolyl substituents were selectively and equally distributed on the 5-hydroxyl functions of terminal- and 2-linked arabinofuranosyl (Araf) residues. Further analysis of the 0-alkylated cell wall through partial acid hydrolysis, NaB[2H]4 reduction, pentadeuterioethylation, and gas chromatography/ spectrometry the mycolyl units unit plished by subjecting fractions to methanolysis in 1 M HC1 in CH,OH at 85 "C for 1 h prior to trimethylsilylation and GS/MS analysis, as described (6). substituents on the cell wall of mycobacteria. The actual proposed structure is demonstrated in order to illustrate the approach. The strategy involved replacement of the mycolyl esters with ethyl ethers and then the determination of the positions of the ethyl groups by formation and analysis of partially acetylated, partially ethylated, and partially methylated alditols. Only the structures of the alditol acetates of residues originally substituted with mycolyl groups

about two-thirds of the available pentasaccharide units are so substituted. Thus, the antigenicity of the arabinan component of mycobacterial cell walls may be explained by the fact that about one-third of the pentaarabinosyl units are not mycolyated and are available for interaction with the immune system. On the other hand, the extreme hydrophobicity and impenetrability of the mycobacterial cell may be explained by the same motif also acting as the fulcrum for massive esterified paraffin residues. New fundamental information on the structure of mycobacterial cell walls will aid in our comprehension of its impenetrability to antibiotics and role in immunopathogenesis and persistence of disease.
The insoluble matrix of the mycobacterial cell wall, namely the product remaining after removal of all soluble proteins, lipids, and carbohydrates, and variously known as cell wall skeleton (1) and cell wall core (2), has been implicated in an array of pathogenic and immunological events associated with tuberculosis and leprosy (3). Following on the fundamental structural work of many over a 50-year period (4), we termed this material the mycolylarabinogalactan-peptidoglycan-protein (mAGPP)l complex and succeeded in defining the major structural features of the arabinogalactan entity (5), of a unique covalent linkage between arabinogalactan and peptidoglycan (6), and of the form and importance of the chemical association between certain proteins and peptidoglycan (7,8).
In particular, out of these more recent efforts arose the recognition of a pentaarabinofuranosyl motif (5) equating the nonreducing termini of the arabinan segment of arabinogalactan with the dominant antigenic determinant of arabinogalactan, if not of the whole mycobacterial cell. Recognition and definition of this entity provided the impetus to address another major structural issue in our quest of a chemical comprehension of the mycobacterial cell wall, namely, the precise location of the mycolate units. It is now demonstrated that these are attached to about two-thirds of the critical nonreducing pentaarabinosyl units, thus presenting an image of a structural arrangement which acts variously as a template for mycolate attachment, and, when not so substituted, as a powerful B-cell immunogen in mycobacterial infections.

Growth of Organisms and Production of mAGPP
Mycobacterium tuberculosis TMC 107 (Erdman), Mycobacterium bovis BCG (Danish strain), Mycobacterium leprae from armadillos, and Mycobacterium smegmatis TMC 607 were prepared in sizable quantity as described previously (5,6). In all cases, mAGPP was prepared as described (5,6); mAGPP is synonymous with the insoluble cell wall matrix.

Methylation of mAGPP
Intact cell walls in some instances were methylated by the base catalyzed procedure of Hakomori (9) as previously described (5). In other situations, cell walls were methylated with methyltrifluoromethanesulfonic acid as described by Prehm (10) with modifications. The method was applied as follows. Five mg of cell walls were suspended in 1 ml of trimethylphosphate (Aldrich) followed by the addition of 300 pl of 2,5-di-tert-butylpyridine, double the usual amount in order to discourage acid-catalyzed cleavage of furanosides, and 100 pl of methyltrifluoromethanesulfonate (both from Aldrich). The reaction mixture was stirred for 6 h at 50 "C, neutralized with pyridine, dialyzed, and freeze dried to recover the methylated mAGPP. Ethylation and concomitant demycolylation was performed as described (5) using CzHsI instead of Cz[zH]51.
Smith Degradation of mAGPP-mAGPP from M. tuberculosis (200 mg) was treated with 10 ml of 0.1 M NaI04 in 50 mM sodium acetate buffer, pH 5.0, and allowed to react for 4 days in the dark at room temperature. Periodate was removed from the insoluble residue by extensive washing with water, and the insoluble residue was reduced by stirring with 130 mg of NaB['HIr in 10 ml of water for 5 h. The reaction mixture was titrated to pH 5 with acetic acid, dialyzed extensively against water, freeze dried, treated with 1 N HCl at room temperature overnight, and the pellet washed with water. Further extraction of the pellet with CHC1, yielded 34 mg of lipid which was applied in CHCl3 to a column (1 X 20 cm) of silica gel followed by a two-step elution regimen of 50 ml each of 2% CHaOH in CHC1, and 6% CH:1OH in CHC1,. Fractions (5 ml) from this column were examined by TLC in CHC13/CHaOH (19:l). Analysis of the products arising from Smith degradation for arabinose and glycerol was accomplished by subjecting fractions to methanolysis in 1 M HC1 in CH,OH at 85 "C for 1 h prior to trimethylsilylation and GS/MS analysis, as described (6).

Application of Replacement 0-Ethylation to Mycolyl Group
Location-Earlier work by others (11-131 had led to the present day concept (14) that cell wall mycolic acids of mycobacteria are attached as esters to position 5 of the terminal arabinosyl residues of arabinogalactan. Amar-Nacasch and Vilkas (15), in particular, had obtained a mycolate of arabinobiose from the cell walls of M. tuberculosis which evidence tended to support the then current idea that the side chains of arabinogalactan consisted of monomycolyltriarabinosyl arrangements. However, in light of the recent recognition (5) of a branched pentaarabinofuranosyl unit ( Fig. 1) as comprising the nonreducing termini of the arabinan component of arabinogalactan, the possibility arose of mycolyl residues being attached at the 5-OH functions of both the t-Araf and the 2linked Araf units; this concept would still adhere to the principle that only the 5-OH functions are acylated.
In order to address these new possibilities, cell walls were methylated with methyltrifluoromethanesulfonate, an acidcatalyzed methylation procedure that does not result in deacylation (10,16). The theoretical outcome of the application of this form of methylation to the pentaarabinosyl motif, this time endowed with a full complement of 5-linked 0-mycolyl residues, is illustrated in Fig The results of this approach as applied to the mAGPP from several species of mycobacteria are shown in Table I Table I). The M. tuberculosis mAGPP complex used in the above analysis contained 35% mycolic acids, according to the weight of CHCh-soluble material obtained from methanolysates, and 25% of the mass consisted of arabinogalactan, as derived from a quantitation of alditol acetates. Calculations based on these percentages, assuming mycolyl substitution on the 5 positions of both the t-Araf and the 2-linked Araf, require that 76% of the total quantities of these 2 residues be mycolylated. The actual data 2,3,5-Tri-0-CH3-5-mono-C2H3-Ara 5-Mycolyl-t-Araf 7.6 * 0.98 a For clarity, only the Ara residues with a free OH group at C-5 (i.e. 2-linked-Araf and t-Araf) are listed. The remaining glycosyl residues, i.e. 5-linked-Araf, 3,5-linked-Araf, 5-linked-Galf, 6-linked-Galf, and 5,6-linked-Galf, were detected as their methylated derivatives with no ethyl substituents and in the proportions originally reported ( 5 ) . Small amounts ( 4 % of the total glycosyl residues) of 5-linked-Araf residues substituted with 0-ethyl groups at C-2 and/or C-3 and 3,5-linked-Araf residues with an 0-ethyl at C-2 were also detected. However, the amounts of these derivatives were variable and were probably due to incomplete methylation; nevertheless, the possibility of small amounts of additional acyl group substitution (either mycolyl or otherwise) on these residues cannot be ruled out. indicated a 67% mycolyl substitution of these 2 residues, in good agreement with theoretical predictions. Also, in the case of all species examined, the amounts of mycolyl-substituted 2-linked Araf and t-Araf were about equal to each other.
Application of Periodate Oxidation to Mycolyl Group Location-Smith degradation could provide independent corroboration of the proposed positioning of the mycolyl residues on AGPP, as demonstrated in Fig. 3. To wit, t-Araf units substi- tuted at C-5 with mycolyl residues should produce [1-2H]3mycolylglycerol upon oxidation, reduction with NaB[2H4], and mild acid hydrolysis (Fig. 3). In addition, the presence of mycolyl groups on the C-5 position of the 2-linked Araf units should be reflected by the appearance of a CHCln-soluble mycolyl-substituted triarabinosylglycerol (Fig. 3). On the other hand, if the 2-linked Araf residues are not so substituted, the resulting triarabinosylglycerol should not be mycolated, i.e. not soluble in CHCl,.
Accordingly, the mAGPP complex from M. tuberculosis was subjected to Smith degradation, and the CHCb-soluble products applied to a column of silica gel which was eluted with increasing concentrations of CH,OH in CHC13. The effluent from the column was monitored by TLC resulting in the recognition of four major fractions (Fig. 4). GC/MS analysis of these after methanolysis and trimethylsilylation showed that fractions A and B contained arabinose and [ l-2H]glycerol ( m / z 294, 205, and 206), thereby confirming the presence of periodate-stable mycolylated arabinosyl residues, i.e. 5-0-mycolyl-2-linked-Araf. 'H NMR of fraction A yielded signals at 6 -0.35 (m) and 6 0.6 (m), arising from cis-cyclopropyl group hydrogens and confirming the presence of cyclopropyl mycolates (17) in this fragment. Also, signals at 6 4.92 (s), 6 4.87 (s), and 6 5.02 (s), arising from the anomeric H-1's of the arabinosyl residues, were evident in fraction A. Deacylation of fraction A, permethylation (9), and GC/MS confirmed the presence of the expected triarabinosylglycerol (Fig. 5). The other two fractions, C and D (Fig. 4), contained [l-2H]glycerol but no arabinosyl residues, thus supporting the evidence for

FIG. 5. GC/MS analysis of the CHC13-soluble triarabinosylglycerol arising from Smith degradation of mAGPP.
Mycolyl groups were concomitantly cleaved and the resulting hydroxyl groups methylated by the base catalyzed methylation procedures (9). A , total ion chromatogram; B , mass spectrum; C, illustration of the source of the major fragment ions.

113-175
periodate labile mycolylarabinofuranosyl residues, i.e. 5-0mycolyl-t-Araf, in the original mAGPP (Fig. 3). Fraction C also co-chromatographed on TLC with authentic 1-0-mycolylglycerol, and 'H NMR analysis showed the same cyclopropyl signal that characterized fraction A ( 6 -0.35, 0.6) as well as two doublet of doublets ( 6 4.23, 12 and 4 Hz; and 6 4.15, 12 and 7 Hz) attributable to the primary H's which used to be attached to C-5 of t-Ara and now are the hydrogens attached to the mycolyl-substituted primary glycerol carbon.
Evidence for the Clustering of Four Mycolyl Groups on Single Pentaarabinosyl Units-The studies described above demonstrated that most of the available terminal Araf and penultimate 2-linked Araf units are mycolylated, and, previously (5), we had shown that these 2 sugar residues invariably occur together as part of the pentaarabinosyl reducing end motif of arabinogalactan (Fig. 1). Nevertheless, several alternative mycolyl substitution patterns could be conceived within the framework of this motif, all consistent with both the methylation and periodate oxidation data; some of these possibilities are illustrated in Fig. 6. In addition, a random distribution of mycolyl substitution was possible.
In order to determine which of these patterns prevailed, mAGPP was methylated followed by replacement of the mycolyl groups with ethyl ethers, as outlined in Fig. 7. From our previous study, we knew that the per-0-methylated terminal pentaarabinosyl unit yielded Fragments IIIa, IVa, and Va when the permethylated arabinogalactan component was subjected to partial acid hydrolysis, NaB[*H], reduction, and pentadeuterioethylation (5). In the present instance, the 5position of the t-Araf and the 2-linked Araf residues should be labeled with CPH6 or CH3 groups, reflecting the presence or absence, respectively, of a mycolyl ester. This form of dual labeling should result in the generation of families (111, IV, and V) of products (Fig. 7), and the relative amounts of the four products within each family should reflect the mycolyl substitution patterns within the intact mAGPP.
The identity of the actual experimentally generated products was determined by GC/MS analysis. The GC retention time and mass spectra of products IIIa, IVa, and Va were already known (5). The remaining three members of each family were identified based on the fact that for every exchange of an ethyl for a methyl group, the retention time of a given alkylated oligoglycosylaliditol is increased. Also, the presence of alkyl groups results in an increase in the m/z value of some but not all of the mass spectral ions in a predictable fashion (Fig. 7).
Selected ion monitoring GC/MS analysis for the expected members of families I11 and IV is shown in Fig. 8A; ions at m/z 303, 317, 331, and 295 were selectively sought and monitored (see Fig. 7 for the origin of these ions). Product IVa was readily recognized from its retention time (16.0 min) and ion profile (m/z 303 and 295). Product IIIa was also recognized from its retention time (16.62 min) and ion profiles (m/z 303 and 295); these products were also identified by a direct comparison of the ion profiles and retention times to those of standards available from previous work (5). Product IVd was readily identified by its retention time (16.60 min; 0.60 min later than IIIa) and its characteristic ion profile ( m / z 295 and 331). Product IIId was recognized by its retention time (17.25 min; 0.65 min later than IIIa) and its characteristic ion profile (m/z 295 and 331). The lack of appreciable amounts of products with a m/z 317 ion profile (Fig. 8A) was indicative of the absence of products IIIb, IIIc, IVb, or IVc. These results combined are consistent with the mycolyl substitution pattern shown in Fig. 6A.
Analysis of the products of family V is shown in Fig. 8B. Product Va, previously characterized (5), was identified based on its retention time of 16 Fig. 7). The generation of products IIIa, IVa, IIId, and IVd (Fig. 8A) and Va and Vd (Fig. 8B) (see also Fig. 7) unequivocally demonstrated that the mycolylated Araf residues are arranged as shown in Fig. 6A. Approximately two-thirds of the terminal nonreducing pentaarabinosyl units (Table I) are mycolylated in this fashion, and consequently about one-third of the terminal pentaarabinosyl units are free of mycolyl residues (Fig. 6A). DISCUSSION The architecture of the cell wall of mycobacteria is central to our understanding of the pressing biological questions of our day such as drug and solute impenetrability (18), antigenicity, and immunoreactivity, notably the issue of antigen processing and presentation by accessory cells (19), immune complex deposition and sequelae (20), and other aspects of immunopathogenesis, such as granuloma formation in tuberculosis (21), and persistence and disease recrudescence in leprosy and tuberculosis. In a spate of intensive fundamental investigations during the period 1950-1970 conducted largely by French (22) and Japanese (23) workers, the fundamental structural features of the bound mycolic acids (17), the peptidoglycan framework (24), and the arabinogalactan heteropolysaccharide (25) were elucidated, and the concept of immunologically active cell wall-bound protein (26) was developed. We recently returned to these topics, in abeyance for 20 years, still without applicable degradative enzymes but in possession of the highly applicable tools of partial depolym- (v) the arabinan chains are attached to the galactan core through the C-5 of some of the 6-linked Galf units (5); and (vi) the galactan of arabinogalactan in turn is linked to C-6 of some of the muramyl residues of peptidoglycan by a special diglycosylphosphoryl bridge with the structure, L-Rhap-( 1-3)-D-GlcNAc-(l+P) (6). Thus, between the old and the new, a considerable body of structural definition has now been conferred on the cell wall core of mycobacteria. The last remaining fundamental structural question concerned the location of the mycolyl residues on the arabinogalactan entity.
With the earlier isolation of ~-arabinosyl-5-mycolate (11-13), the terminal segments of arabinogalactan clearly became implicated as the primary binding site for cell wall mycolates. However, these observations predated the recent surge of knowledge on the primary structure of arabinogalactan and, in particular, the recognition of a pentaarabinofuranosyl arrangement as representing the majority of the nonreducing termini of the arabinan component of arabinogalactan. Present evidence now clearly states that these units are fully  (Fig. 7). All eight compounds yielded m / z 295 (see Fig. 7). terms of a phalanx of parallel hydrocarbon chains (Fig. 9). It is also clear, however, that not all pentaarabinosyl units are mycolylated; methylation analysis (Table I) suggests that on average only about two-thirds of the available pentaarabinosyl units are so substituted. We are also struck by the evidence, arrived at from methylation analysis and periodate oxidation, that mycolylation is an all-or-none event, in other words, full tetra-0-mycolylation or no mycolylation prevails. This striking fact, combined with prior evidence (5) for the existence of several arabinan chains emanating from a single linear galactan segment, further suggests that within the native peptidoglycan-bound mycolylarabinogalactan there may exist both mycolylated and nonmycolylated arabinan chains. Perhaps the latter, in combination with the arabinan segments of lipoarabinomannan (30)3, also nonmycolylated, provide the primary epitopes of antibody response and contribute as the long sought conduits for solute transport through mycobacterial cell wall. Finally, our new image (Fig. 9) of the essence of mycobacterial cell walls may at last provide a rational basis of acid fastness (32) and its loss during death of mycobacteria due to chemotherapy (31).