Novel O-methylated terminal glucuronic acid characterizes the polar glycopeptidolipids of Mycobacterium habana strain TMC 5135.

Mycobacterium “habana” strain TMC 5135, which has been proposed as a vaccine against both leprosy and tuberculosis, is considered to be a strain of serotype I of the recognized species Mycobacterium simiae. We have now shown that each of these strains possesses characteristic polar glycopeptidolipids (GPL) which are sufficiently different to allow unequivocal strain identification. Thin layer chromatographic analysis demonstrated that M. habana synthesizes a family of apolar GPLs and three distinct polar GPLs (pGPL-I to -III) which exhibited migration patterns different from those of M. simiae serotype I (pGPL-Sim). Using a combination of chemical, mass spectrometric, and proton-NMR analyses, the GPLs from M. habana were determined to be based on the same generic structure as those from the M. avium complex, namely N-fatty acyl-D-Phe-(O-saccharide)-D-allo-Thr-D-Ala-L-alaninyl-O-monosaccharide. The de-O-acetylated apolar GPLs contain a 3-O-Me-6-deoxy-Tal attached to the allo-Thr and either a 3-O-Me-Rha or a 3,4-di-O-Me-Rha attached to the alaninol. In the pGPLs, oligosaccharides were found to be attached to the allo-Thr. The oligoglycosyl alditol reductively released from the least polar pGPL-I was fully characterized as L-Fucpα13-(6-O-Me)-D-Glcpβ13-(4-O-Me)-L-Rhapα13-L-Rhapα12-(3-O-Me)-6-deoxy-Tal. In pGPLII and -III, the terminal Fuc residue is further 3-O-methylated and 4-O-substituted with an additional 2,4-di-O-Me-D-GlcA and 4-O-Me-D-GlcA, respectively. The corresponding oligosaccharide from pGPL-Sim was shown to be of identical molecular weight to pGPL-II but terminating with a 3,4-di-O-Me-GlcA. Enzyme-linked immunosorbent assay-based serological studies using anti-M. habana and anti-M. simiae sera against whole cells and purified pGPLs firmly established the polar GPLs as important antigens and indicated that the terminal epitopes L-Fuc-, 2,4-di-O-Me-D-GlcA, and 4-O-Me-D-GlcA uniquely present in pGPL-I, -II, and -III, respectively, confer sufficient specificity for the identification of M. habana as a distinct serotype of M. simiae.

Leprosy and tuberculosis are grave diseases with large worldwide prevalences (Raviglione et al., 1995;Noordeen, 1994). If leprosy seems, for the present and at considerable cost, likely to be controlled by effective and available chemotherapy, tuberculosis incidence is rising and infections with novel multiple drug-resistant strains are effectively incurable (Bloom and Murray, 1992). Although BCG is a highly effective vaccine against both diseases in some populations, in others it is without detectable effect (Fine, 1995), and alternative vaccines are being sought. The need for novel tuberculosis vaccines is particularly urgent.
Mycobacterium habana strain TMC 5135 has been used successfully to vaccinate mice against tuberculosis (Gupta et al., 1979) and leprosy (Singh et al., 1985;Singh et al., 1989) and has been proposed as a human vaccine against both diseases (Singh et al., 1991). This strain was one of a group of 35 similar strains isolated in Havana from patients with lung disease (Valdivia-Alvarez et al., 1971). Although the symptoms resembled pulmonary tuberculosis, the M. habana strains differed markedly from Mycobacterium tuberculosis itself. It was originally proposed as a novel mycobacterial species, but subsequent taxonomic investigation using classical "biochemical" tests and seroagglutination indicated that M. habana strains did not differ sufficiently from serotype I of Mycobacterium simiae, a recognized mycobacterial species, to deserve the status of a separate species (Meissner and Schröder, 1975). However, these authors found that sera raised against M. habana, absorbed with M. simiae, retained some ability to agglutinate M. habana strains, indicating that these possess antigenic determinants absent from M. simiae. A similar conclusion was reached on the basis of reciprocal skin tests in guinea pigs and DNA hybridization measurements (Baess and Magnusson, 1982).
Most species of mycobacteria form clumps spontaneously during growth so that seroagglutination techniques cannot be used to identify them. However, a few species, notably members of the so-called Mycobacterium avium-Mycobacterium intracellulare complex but also some others, can be successfully typed by seroagglutination (Schaefer, 1967). In these species the dominant surface antigens are the distinctive polar glycopeptidolipids (GPLs) 1 (Brennan, 1988). These are members of a larger family of GPLs, all related to "mycoside C" (Jollès et al., 1961), which also includes apolar types devoid of significant serological activity. The polar types are distinguished by the possession of complex oligosaccharides which are the major antigenic determinants (Yanagihara et al., 1985). For strains of mycobacteria producing such polar GPLs, definitive identification is simplified and may be effected by TLC, ELISA, or physicochemical analysis (Denner et al., 1992), as well as the original seroagglutination methods.
In this paper, we have investigated the presence and structures of GPLs in M. habana TMC 5135. In addition to a family of simple apolar GPLs, we found three polar GPLs carrying complex oligosaccharides with unique sequences. Similar structures were identified in the polar GPLs of M. simiae serotype I but were clearly distinguishable from each of those of M. habana by the degree and positions of O-methylation on the nonreducing terminal glucuronic acid residue. Serological studies provided further insights into the known cross-reactivities, which can now be structurally rationalized to some extent. Identification of these unique molecules in M. habana TMC 5135 provides a simple serological or chromatographic method of defining the strain, which will be essential for quality control if it is to be used to protect humans against tuberculosis or leprosy. It also underlines the biochemical versatility of mycobacteria, and the strong effect of apparently minor variations in chemical structure on serological activity/specificity. Extraction and Purification of GPLs-For lipid production, mycobacteria were grown on slopes of Middlebrook 7H9 medium with OADC supplement (Difco) in 50-ml screw-capped polystyrene jars (R.S. Jars) for 4 -6 weeks. The cultures were sterilized by 500 krad of ␥ radiation from 60 Co before harvesting. Cells were washed off the surface of the medium by vortexing with water, washed once with water, and freezedried. Dry mycobacterial mass was then extracted with ethanol/ether (1:1 by volume) at room temperature for 24 h, followed by chloroform/ methanol (2:1 by volume) under similar conditions. Solvent extracts were dried in a rotary evaporator, then deacetylated by the method of Brennan and Goren (1979). Apolar and polar GPL were isolated by preparative TLC on 20 ϫ 20-cm Silica Gel 60 plates (Merck), using chloroform/methanol (9:1 by volume) and chloroform/methanol/water (65:25:4 by volume), respectively, as solvents. Glycolipids were identified in side bands cut from the plates, using ␣-naphthol-sulfuric acid, and corresponding iodine vapor-stained bands on the main part of each chromatographic plate were extracted from scraped off silica gel with chloroform/methanol (2:1 by volume) or chloroform/methanol/water (65: 25:4 by volume). They were repurified by TLC until homogeneous. Identical methods of growth and processing were used to obtain polar GPLs from M. simiae serotype I, and, on a smaller scale, from the related strains. Polar GPL from M. avium serovar 19 was prepared as described before (Chatterjee et al., 1988).

Mycobacteria
Analytical TLC-Lipids were analyzed by TLC using Silica Gel 60 TLC and high performance TLC plates (Merck), and chloroform/methanol (9:1 by volume) and chloroform/methanol/water (65:25:4 by volume) as developing solvents for separating apolar and polar GPL, respectively. Lipids were detected with dichromate/sulfuric acid (general reagent for lipids), ␣-naphthol sulfuric acid (reagent for glycolipids) and exposure to chlorine gas followed by spraying with starch-potassium iodide (reagent for amine and imine groups).
Release of the Oligoglycosyl Moiety-The oligoglycosyl moiety assumed (and subsequently demonstrated) to be attached to allo-Thr was released from GPLs by reductive elimination (500 l of 10 mg/ml sodium borohydride, 0.5 M sodium hydroxide, 500 l of ethanol, stirred at 60°C for 16 h) as oligoglycosyl alditol and designated as OSE. The sample was then decationized directly with Dowex (50W X-8(H)) beads, centrifuged, and the supernatant repeatedly co-evaporated to dryness with 10% acetic acid in methanol under nitrogen.
Compositional Analysis-Amino acids in the GPLs were identified using a Beckman 6300 amino acid analyzer after hydrolysis of the glycolipids in 4 M HCl at 110°C for 18 h in sealed tubes, and also by GC and GC-MS analysis of their heptafluorobutyryl derivatives (MacKenzie and Tenaschuk, 1975) to determine the absolute configurations as described previously (Besra et al., 1993). Briefly, the N(O)-heptafluorobutyryl esters were prepared by hydrolysis in 6 N HCl at 110°C for 12 h, followed by treatment with 100 l of 3 N HCl in (R)-(Ϫ)-2-butanol at 120°C for 20 min, and then 100 l of anhydrous ethyl acetate and 40 l of heptafluorobutyric anhydride at 150°C for 5 min. The DL-configurations of the O-methylated glycosyl residues were determined by GC-MS analysis of the trimethylsilyl derivatives of (R)-(Ϫ)-and (S)-(ϩ)-2-butyl glycosides (Gerwig et al., 1978) after initial de-O-methylation of the permethylated OSE with 1 M boron tribromide in dichloromethane at Ϫ80°C for 30 min followed by 16 h at room temperature (Chatterjee et al., 1987). For GC and GC-MS analysis of glycosyl composition and linkage positions, partially methylated alditol acetates were prepared from purified GPLs and perdeuteromethylated oligoglycosyl alditols, respectively, by hydrolysis in 2 M trifluoroacetic acid at 120°C for 2 h, reduction with sodium borodeuteride, and peracetylation. The terminal uronic acid was similarly identified as partially O-CH 3 -, O-CD 3 -, and O-Ac-hexitol by GC-MS linkage analysis after reduction of the perdeuteromethylated OSE with lithium aluminium hydride in tetrahydrofuran at 80°C for 2 h (Chatterjee et al., 1987).
GC of the alditol acetates was performed on a Hewlett-Packard 5890 instrument fitted with a DB-23 capillary column (15 m, 0.25-mm inner diameter, 0.25 m film thickness; J & W Scientific) using a temperature gradient of 150°C for 4 min and then 2°C/min to 190°C. GC-MS was carried out on a Hewlett-Packard 5890 Gas Chromatograph connected to a Hewlett-Packard 5790 Mass Selective Detector. Partially methylated and trideuteromethylated alditol acetates were analyzed on a SP 2380 capillary column (30 m, 0.25-mm inner diameter, 0.20-m film thickness; Supelco) using a temperature gradient of 30°C/min to 190°C and then 4°C/min to 250°C. The trimethylsilyl derivatives of butyl glycosides were injected at 80°C and chromatographed on a DB-1 capillary column (J & W Scientific) with a temperature gradient of 30°C/min to 100°C and then 10°C/min to 265°C.The heptafluorobutyryl derivatives of amino acids were resolved with a temperature gradient of 30°C/min to 100°C and then 8°C/min to 250°C.
Derivatization and FAB-MS Analysis-Intact GPLs were analyzed by FAB-MS either in their native state, or after standing in methanolic-HCl for a suitable period of time, or after perdeuteromethylation/perdeuteroacetylation as described previously Besra et al., 1993). OSE samples were similarly perdeuteromethylated and perdeuteroacetylated. FAB-mass spectra were obtained using a VG Analytical ZAB-2SE 2FPD or an Autospec mass spectrometer fitted with a cesium ion gun operated at 25-30 kV. Data acquisition and processing were performed using the VG Analytical Opus® software. Both monothioglycerol and m-nitrobenzyl alcohol were used as matrices to optimize for different fragmentation pathways.
1 H NMR Analysis-The GPLs were deuterium exchanged with deuterated methanol and dissolved in CDCl 3 (100%) and CD 3 OD (100%) in the ratio of 4:1 (v/v). The oligoglycosyl alditols were deuterium exchanged and dissolved in 500 l of deuterium oxide (100%) atom. 1 H-Detected two-dimensional [ 13 C-1 H] shift correlation (heteronuclear multiple-quantum correlation) spectra were recorded with a Bruker AM-500 spectrometer using an inverse detection probe, as described by Bax and Summers (1986).
Serological Studies-Antisera were raised in rabbits to M. habana TMC 5135 and M. simiae serotype I by a method based on that of Anz et al. (1969), except that phenol was omitted from the suspending buffer and the rabbits were vaccinated in multiple sites. The vaccines consisted of 3.2 mg (dry mass) of mycobacteria suspended in 20 ml of buffer and heated at 80°C for 30 min. For the initial vaccination the suspensions were mixed 1:1 (by volume) with Freund incomplete adjuvant and injected subcutaneously (0.25 ml into each of six sites). After 14 and 28 days, two booster doses of 0.25-ml suspensions without adjuvant were given. Test bleeds were taken 7 days after the last vaccination and since these were found to contain satisfactory levels of antibody the animals were exsanguinated and sera prepared from the whole blood.
Cross-reactivities among the whole cells and the polar GPLs against the antisera were examined by ELISA. Mycobacteria were applied to ELISA plates (Nunc Maxisorp) as water suspensions, 4 g (dry mass) in 0.1 ml, and dried on to the plates at 37°C. Glycolipid extracts or purified polar GPLs were suspended in water by vigorous vortexing or by sonication, applied to ELISA plates at 0.1 ml/well (containing 1-4 g of lipid) and dried onto the plates at 37°C. Plates were blocked with 1% dried milk powder (Marvel) in phosphate-buffered saline, the antisera were used at 1/25 or greater dilutions, the second antibody was peroxidase-conjugated swine anti-rabbit immunoglobulin (Dako) used at 1/1000, the color developer was ABTS (Sigma) and absorbance of wells were measured at 405 nm in a Biotek EL312 plate reader. Tween 20 was not included in any of the washing solutions or diluents when glycolipids were being used as antigens. For cross-absorption experiments with pGPL-I to -III and pGPL-Sim, glycolipid suspensions at 40 g/ml were mixed with the diluted sera (1/100 dilution) for 2 h at 37°C before adding 0.1 ml of the serum to wells containing 3.5 g of glycolipid. For experiments involving GPL from M. avium serotype 19, sera at 1/200 dilution were absorbed with 20 and 40 g/ml of glycolipid suspension for 66 h before being added to wells containing 1 g of glycolipid.

RESULTS
The GPLs of M. habana-TLC analysis showed that M. habana possessed a family of apolar GPLs, a GPL of intermediate polarity (pGPL-I) and two very polar GPLs (pGPL-II and III) (Fig. 1). The distribution of polarities of these glycolipids was comparable to that reported for the two archetypal GPL families (polar and apolar) of the M. avium complex (Brennan, 1988). The apolar components and two of the polar lipid components, pGPL-I and pGPL-II, were largely extracted from the dry bacteria by ethanol/ether (1:1, v:v), whereas pGPL-III was found also in the chloroform/methanol (2:1, v:v) extract. GC and GC-MS analysis of the amino acid composition, in comparison to authentic standards, showed that both the apolar and polar GPLs contained D-Ala, D-Phe, D-allo-Thr, and L-alaninol in approximately equimolar amounts, indicating that these GPLs might share the tetrapeptide core sequence found in all mycobacterial GPLs characterized so far except those from Mycobacterium xenopi (Aspinall et al., 1995;López-Marin et al., 1992Khoo et al., 1995;Rivière et al., 1993).
Apolar GPLs-Mild deacylation reduced the mobilities of all the bands of apolar GPL observable on chromatograms, indicating that these substances were O-acetylated in their native state. The deacylated apolar GPLs could be resolved into at least 5 major bands on TLC, which were isolated and designated aGPL-I to -V (Fig. 1B). A number of minor apolar components were present, but these were not further investigated. The glycosyl compositions of the five aGPLs isolated fell into 2 groups: the more mobile species (aGPL-I to -III) contained 3-O-Me-6-deoxy-Tal and 3,4-di-O-Me-Rha whereas the less mobile ones (aGPL-IV and -V) contained 3-O-Me-6-deoxy-Tal and 3-O-Me-Rha. As described previously in the FAB-MS sequencing of the apolar GPLs from Mycobacterium senegalense (Besra et al., 1994) and Mycobacterium butyricum (Khoo et al., 1995), the perdeuteromethylated derivatives afforded diagnostic fragment ions which allowed definitive assignment of the tetrapeptide sequence and the identity of the two monosaccharide residues attached to allo-Thr and alaninol, respectively. From the m/z values of these (Table I) and the molecular ions (Table II), it could be concluded that: 1) each of the bands isolated by TLC comprised a mixture of molecular species of aGPL whose het-erogeneity resided in their N-fatty acyl chains. The molecular ions and all fragment ions containing the fatty acyl chains comprised distinctive clusters of signals dominated by two or more major species. In contrast, fragment ions which did not contain the fatty acyl chains were characterized by single signals.
2) The monosaccharide attached to allo-Thr was invari-

STRUCTURE I
Polar GPLs-Chromatographic mobilities of the polar GPLs were unaffected by mild deacylation, implying that O-acetyl groups were not present. As with the apolar GPLs, their amino acid composition indicated a conventional N-acyl-D-Phe-D-allo-Thr-D-Ala-L-alaninol core. However, glycosyl composition analysis clearly showed that the polar GPLs possessed a much more complex saccharide content. GC and GC-MS analysis of the alditol acetates derived from pGPL-I showed the presence of 6 distinct sugar residues in approximately equimolar amounts, namely 3,4-di-O-Me-Rha, 3-O-Me-6-deoxy-Tal, 4-O-Me-Rha, Rha, Fuc, and a 6-O-Me-Hex. pGPL-II and -III had a common alditol acetate composition which differed from that of pGPL-I in having an additional 3-O-Me-dHex and a smaller amount of Fuc. Oligoglycosyl alditols, referred to as OSE-I, -II, and -III, were released from each of the pGPLs by reductive elimination and fully characterized as follows.
OSE-I from pGPL-I-The FAB-mass spectrum of the perdeuteroacetyl derivative of OSE-I afforded an (M ϩ Na) ϩ molecular ion at m/z 1326 together with prominent A-type oxonium ions at m/z 282, 548, 753, and 989, thus establishing the sequence as shown in Structure II.
This sequence was further corroborated by the perdeuteromethylated derivative which yielded an [M ϩ Na]  revealed (i) four distinct anomeric protons at ␦5.42 (J 1,2 ϭ 3.5 Hz), 5.25, 5.15 (J 1,2 ϳ 1.0 Hz, indicative of a anomers in the manno configuration) and 4.85 (a doublet, J 1,2 ϭ 7.7 Hz, indicative of a ␤-hexosyl unit); (ii) three proton singlets between ␦3.52-3.75 assignable to three O-Me groups; and (iii) four sets of doublets arising from H-6 of the 6-deoxyhexoses, suggesting the presence of 4 different 6-deoxyhexoses. When further analyzed by two-dimensional 1 H-1 H correlated spectroscopy (COSY), the four anomeric protons of unit intensity and their corresponding H1-H2 cross-peaks were observed in the lowfield region of the spectrum. The two cross-peaks at 1 ϭ 5.09 ppm, 2 ϭ 4.31 ppm (J 1,2 ϳ 1.2 Hz), and 1 ϭ 4.99 ppm, 2 ϭ 4.09 ppm (J 1,2 ϳ 1 Hz) confirmed the presence of two deoxyhexoses in their manno configuration, i.e. the 4-O-Me-Rha and Rha, whereas the most downfield cross-peak at 1 ϭ 5.35 ppm and 2 ϭ 3.85 ppm (J 1,2 ϳ 3.5 Hz) indicated the presence of a terminal ␣-Fuc residue. Finally, the cross-peaks at 1 ϭ 4.71 ppm, 2 ϭ 3.59 ppm (J 1,2 ϳ 7.5 Hz) established that the 6-O-Me-Glc was in the ␤-anomeric configuration (Table III). Additional information on the anomeric configuration of the 3-O-Me-6-deoxy-Tal and 3,4-di-O-Me-Rha was obtained from NMR analysis of the parent pGPL-I using the 1 H detected two-dimensional [ 1 H, 13 C]heteronuclear multiple quantum correlation experiment ( Fig. 2A) and measured the J C 1 H 1 values by inverse-detection two-dimensional-NMR. Thus the proton resonating at ␦5.05 was assigned to the C-1 at ␦101.6 (J 1,2 ϭ 3.5 Hz, from the 1 H NMR and J C 1 H 1 ϭ 172 Hz) and the 1 H-1 H connectivity of H-2, H-3, H-4, H-5, and H-6,6Ј identified the glycosyl residue to be an ␣-Fuc residue. The anomeric protons resonating at ␦4.99, 4.93, 4.68 have C1s resonating at ␦101.65, 102.2, and 96.69, respectively. The broad singlet at ␦4.93, which appeared to be a single proton, could be further resolved TABLE I Selected fragment ions observed in the FAB-mass spectra of perdeuteromethylated aGPLs A full set of complementary fragment ions were observed but only ions critical to the structural analysis are tabulated here. The first set of fatty acyl-containing ions afforded the peptide sequence as well as defining the mono-O-Me-dHex attached to the allo-Thr in each of the aGPLs. The second set of ions confirmed that the dHex residue attached to alaninol is either mono-or di-O-methylated. *Thr and Alaninol* represent glycosylated allo-Thr and alaninol, respectively. ⌬Thr represent the allo-Thr residue after eliminating its saccharide moiety.  m/z values calculated from the molecular ions after subtracting the known mass of the peptide and saccharide moieties. b m/z values calculated from the ion clusters attributed to N-fatty acyl-Phe-CO ϩ ( Table I). The presence of free OH group(s) can be determined from the total number of trideuteromethyl group(s) incorporated in the perdeuteromethylated derivatives, with each trideuteromethyl giving an increment of 17 units. The additional 3-unit increment observed for m/z 521/535 indicated the possible presence of a methyl group which can be replaced by a trideuteromethyl under the methylation conditions. STRUCTURE II into two sets of cross-peaks in both the 1 H-1 H COSY and the heteronuclear multiple-quantum correlation carbon spectra, indicating a total of 4 related anomeric signals. The J C 1 H 1 values for each of these protons were calculated to be 173.5, 173.5, 173, and 170.5 Hz respectively, thus established that all three Rha and the 6-deoxy-Tal in pGPL-I were ␣-anomers.
OSE-II and -III-The monosaccharide composition as revealed by alditol acetate analysis suggested that pGPL-II and -III might carry an identical oligoglycosyl moiety which differed from that of pGPL-I in having a 3-O-Me-dHex substituted for the terminal Fuc. Indeed, GC-MS linkage analysis of OSE-II and -III gave the same total ion current chromatogram. Comparison with that of OSE-I indicated that the peak attributed to terminal Fuc had disappeared concomitant with the appearance of a new peak (R T 14.07 min, ions at m/z 121, 143, and 203, eluting at the same time as a 4-linked Fuc standard) identified as 4-linked 3-O-Me-Fuc. Closer examination showed that a small amount of non-O-methylated 4-linked Fuc was also present as evident from the weaker ions at m/z 146 and 206. Significantly, no terminal sugar was detected.
FAB-MS analysis of the perdeuteroacetylated OSE-II afforded two major molecular ions at m/z 1436 and 1499 together with a pair of less abundant ions at m/z 1467 and 1530. A mild methanolic-HCl treatment of the sample followed by re-deuteroacetylation yielded one molecular ion at m/z 1513 accompanied by a weak signal at m/z 1544. These data were consistent with OSE-II being a mixture of major and minor constituents each of which carried a single carboxyl group. Partial lactonization during deuteroacetylation  afforded the pairs of signals separated by 63 mass units. Treatment with methanolic-HCl converted both the free carboxyl group and the lactone to the methyl ester. To determine the identity of the (O-Me) 2 HexA, the perdeuteromethylated sample was reduced with lithium aluminium hydride and subjected to GC-MS linkage analysis. In addition to the peaks already identified, a new peak was observed, eluting at a retention time similar to a 6-linked Glc standard. This was identified as a 6-linked 2,4-di-O-Me-Glc based on the ions observed (m/z 105, 118, 129, 165, 176, 189, and 236). The terminal residue of OSE-II was thus established as a 2,4-di-O-Me-GlcA 4-linked to a predominantly 3-O-methylated Fuc. Taking all the data into account, it could be concluded that OSE-II shared the same core sequence with OSE-I but differed at the nonreducing terminus, where the Fuc in OSE-II was extended with a 2,4-di-O-Me-GlcA at position 4, and was largely O-methylated at position 3.
Linkage analysis on lithium aluminium hydride-reduced OSE-III gave the same total in current profile as that from OSE-II. However, the electron impact-mass spectrum of the 6-linked Glc peak afforded ions at m/z 108,121,129,168,176,189, and 236 consistent with a 1,5,6-O-Ac-4-O-Me-2,3-di-O-d 3 -Me-hexitol. Thus, OSE-III appeared to differ from OSE-II in having a 4-O-methylated glucuronic acid instead of a di-Omethylated residue. FAB-MS analysis of the perdeuteroacetylated and perdeuteromethylated derivatives confirmed this conclusion. The former yielded molecular ions 31 mass units higher than those of OSE-II (both before and after methyl esterification) while the latter gave molecular ions 3 mass units higher, fully consistent with OSE-III having one O-Me group fewer than OSE-II. In addition, the esterified perdeuteroacetylated OSE-III gave a prominent A-type ion at m/z 295, corresponding to a methyl esterified, mono-O-methylated HexA (cf. ion at m/z 264 for OSE-II). The difference in O-methylation between OSE-II and -III was therefore localized to the terminal HexA, corroborating the linkage analysis data. Two-dimensional 1 H-1 H COSY NMR analysis of OSE-II revealed 4 distinct anomeric protons with corresponding crosspeaks for H-2 in the low field region (Fig. 2B, Table III) which were similar to those of OSE-I. An additional anomeric proton (somewhat obscure due to the residual water peak) with an intense cross-peak for H-2 ( 1 ϭ 4.82 ppm, 2 ϭ 3.65 ppm; J 1,2 ϭ 7.3 Hz) was detected in the spectrum, indicative of an additional ␤-hexosyl residue not present in the spectra of pGPL-I and OSE-I. The cross-peak connectivity of this residue lead to the assignment of H-3, H-4, H-5, but no other methylene protons or deoxy protons were observed. It was concluded that this additional glycosyl residue was not a 6-deoxy-Hex nor a Hex but was almost certainly attributable to di-O-Me-GlcA, occurring as a ␤-anomer.
In summary, OSE-I from pGPL-I was firmly established as L-Fucp␣133-  but was only mono-O-methylated at position 4 of the GlcA, consistent with it being the most polar of all the GPLs characterized. After lithium aluminium hydride reduction followed by de-O-methylation, only D-Glc, L-Rha, and L-Fuc were detected in subsequent GC-MS analysis of the trimethylsilylated R-(Ϫ)-2-butyl glycoside derivatives of OSE-II and -III, indicating a D-configuration for the additional terminal GlcA. FAB-MS Analysis of pGPLs-FAB-MS analysis of native pGPL-I afforded a cluster of (M ϩ Na) ϩ molecular ions which could be categorized into three major groups: (i) m/z 1821/ 1835; (ii) 1853/1867; and (iii) 1899/1913, reflecting the heterogeneity associated with the N-fatty acyl chains. As observed previously (Besra et al., 1993;Khoo et al., 1995), the pGPLs readily yielded protonated ions corresponding to the lipopeptide cores, especially when treated with mild methanolic-HCl. A comparison of the masses of these ions with the molecular ions for intact GPL enabled assignments of the compositions of the saccharide appendages. These compositions were consistent with those deduced from the data on OSE-I. More definitive sequencing of the intact pGPL-I was afforded by the key frag- FIG. 2. A, 1 H detected [ 1 H, 13 C]heteronuclear multiple-quantum correlation spectrum of pGPL-I; cross-peak labels correspond with those on the structure shown. The residues 3 and 4 are inter alia. B, two-dimensional, 500 MHz 1H-1H COSY spectrum of OSE-II. ment ions produced by the perdeuteromethyl derivatives (Fig.  4), as illustrated in Structure III. The data unambiguously established the peptide sequence as conforming to the conventional N-acyl-Phe-Thr-Ala-alaninyl-di-O-Me-dHex core, as well as providing evidence for the attachment of a pentasaccharide (OSE-I) directly to the allo-Thr residue.
In agreement with the conclusion drawn from the analysis of the OSEs, the ion at m/z 1431 afforded by pGPL-I was shifted to m/z 1649/1652 and 1652/1655 in the FAB-spectra of perdeuteromethylated pGPL-II and -III, respectively, consistent with the presence of an additional HexA residue in the latter two pGPLs. In contrast, the key fragment ions corresponding to N-acyl-Phe-CO ϩ , and all others which had eliminated the oligosaccharide appendage, remained unchanged. It was therefore concluded that the only structural difference among the pGPLs resided in the nonreducing termini of their oligosaccharide appendages as defined by analysis of the released OSEs. All shared the same peptide core structure as well as a similar distribution of fatty acyl chains. As with the aGPLs, the fatty acyl composition could be inferred from the mass data which indicated the presence of three series as follows: (i) fatty acyl groups giving acylium ions of m/z 443/457 and having no free OH groups; (ii) fatty acyl groups giving acylium ions of m/z 475/489 which could incorporate 3 deuteromethyl groups; and (iii) fatty acyl groups giving acylium ions of m/z 521/535 which were therefore similar to those of aGPL-I and -IV.
Polar GPLs of M. simiae-TLC demonstrated two major polar GPLs in extracts of M. simiae serotype I strain 26110, migrating close together on the plate (Fig. 1C). The more abundant of the two had a slightly greater mobility. Both of their mobilities were greater than pGPL-II of M. habana but less than pGPL-I. Additional strains of M. simiae serotype I were grown on a small scale and the glycolipids extracted by ethanol/ ether (1:1 by volume) were examined by high performance TLC. Glycolipid extracts from strain 26111, identified as "normal" M. simiae serotype I, displayed the same pattern as strain 26110. However, glycolipid extracts from strains 26112 and 26113, both identified as "habana like," displayed the distinctive pGPL pattern of M. habana TMC 5135.
To assess rapidly the structural determinants on the polar GPLs from true M. simiae serotype I (strain 26110) (designated pGPL-Sim) which would account for their polarity and distinguish them from those of M. habana, the oligoglycosyl alditol from the pooled substances (OSE-Sim) was reductively elimi-nated from pGPL-Sim and analyzed by FAB-MS after derivatization. The perdeuteromethylated OSE-Sim afforded a major (M ϩ Na) ϩ molecular ion at m/z 1236 accompanied by a weak signal 3 mass units lower, whereas the perdeuteroacetylated sample afforded a major (M ϩ 2Na-H) ϩ molecular ion at m/z 1521 together with a weaker signal 31 mass units lower (therefore containing one extra O-Me). Treatment of the latter sample with methanolic-HCl followed by re-deuteroacetylation yielded (M ϩ Na) ϩ molecular ions at m/z 1482 and 1513. Thus, by analogy with results from OSE-II and III, the data indicated that OSE-Sim might also contain a terminal hexuronic acid which was susceptible to methyl esterification. The m/z value of the major perdeuteroacetylated species (1513 after esterification) is identical to that of OSE-II. Prominent A-type ions were observed at m/z 264 and 232, the former corresponding to a methyl esterified di-O-Me-HexA ϩ . The apparent loss of 32 units from m/z 264 to give the ion at m/z 232 which was not found in the spectrum of OSE-II suggested that one of the O-Me on the HexA was likely to be at position 3.
While the minor component of OSE-Sim (released from a pool of both pGPL bands) containing an extra O-methyl group was not detected in OSE-II, the major species appears to be of similar composition to OSE-II, as demonstrated by the m/z values of both the derivatives. However, linkage analysis on the perdeuteromethylated sample showed that OSE-Sim is not identical to OSE-II. Three major peaks were identified: Thus, there was an indication that there were antigenic differences between the strains and that the whole cells of these two strains might readily be distinguished on the basis of their differential reactivities with the two antisera.
Whole ethanol/ether extracts containing GPLs were also well recognized by the antisera (results not shown), but the crossreactivity was not fully reciprocal. It was observed that anti-M. simiae serum recognized M. habana glycolipid extracts quite well but the opposite reaction, i.e. anti-M. habana serum against M. simiae glycolipids, was weak and only significant a thigh serum concentrations. Using purified pGPLs, it could be further shown that, in addition to homologous recognition, anti-M. simiae serum recognized a mixture of pGPL-II and pGPL-III but not pGPL-I from M. habana, whereas anti-M. habana serum did not recognize pGPL-Sim from M. simiae (Fig. 5).
Absorption of antisera with purified GPLs before adding them to the ELISA plates, using a similar amount of GPLs for absorption as was coated on to the ELISA well, gave results as predicted from the above observations but with one notable anomaly (Fig. 6). Absorption of anti-M. simiae serum with pGPL-Sim reduced both homologous reactivity and reaction against pGPL-IIϩIII of M. habana. However, absorption of this serum with pGPL-IIϩIII reduced only the reaction against pGPL-IIϩIII. These effects were statistically significant (p Ͻ 0.01 in a 2-way t test).
Taken together, the serological studies suggested that pG-PLs were important antigenic determinants in these strains. Further evidence in support of this observation was obtained by extending the serological studies to polar GPL from M. avium serotype 19 (pGPL-19) which shares a very similar structure with pGPL-II, -III, and -Sim, particularly in containing the terminal 3,4-di-O-Me-␤-D-GlcA (Chatterjee et al., 1988). As expected, pGPL-19 was recognized by both anti-M. habana and anti-M. simiae sera (results not shown). In the case of anti-M. simiae serum, the reaction was similar in magnitude to the reaction with the homologous pGPL-Sim and greatly reduced when the anti sera was pre-absorbed with pGPL-19. On the other hand, reactivity of pGPL-19 with anti-M. habana was readily detectable but only at about 25% of the magnitude of reaction against anti-M. simiae serum, and declined much more rapidly with increasing serum dilution than did the reaction with homologous pGPL-I and pGPL-II. DISCUSSION The polar glycopeptidolipids of the M. avium complex were the first surface antigens used to identify mycobacterial strains (Schaefer, 1967;reviewed by Brennan, 1988). Species or strainspecific glycolipids structurally based on the same prototype mycoside C have since been identified in other mycobacterial species such as Mycobacterium peregrinum, Mycobacterium senegalense, Mycobacterium porcinum, and Mycobacterium butyricum (López-Marin et al., 1991Besra et al., 1994;Khoo et al., 1995) although in these species the GPLs are mainly of the apolar types without the haptenic oligosaccharide appendages which characterize the M. avium serotypes.
The GPLs isolated from M. habana strain TMC 5135 and characterized in this study closely resemble those of the M. avium complex in, 1) the core structure of fatty acyl-D-Phe-(Osaccharide)-D-allo-Thr-D-Ala-L-alaninyl-O-monosaccharide; 2) the long fatty acyl chain length averaging at C-30; 3) the presence of apolar and polar subclasses, both with a single 3-O-methyl or a 3,4-di-O-methyl-Rha residue at the alaninol end; and 4) complex oligosaccharide appendage attached to allo-Thr via a 3-O-Me-6-deoxy-Tal in the case of the polar GPLs. The 4-O-Me-or 2,4-di-O-Me-GlcA at the nonreducing termini of the more polar pGPL-II and -III represent novel terminal epitopes not previously identified on mycobacterial surface glyolipids (Aspinall et al., 1995) although they closely resemble the terminal 3,4-di-O-Me-GlcA on the GPL of M. avium serovar 19 (Chatterjee et al., 1988). Interestingly, the polar GPLs of M. simiae serotype I strain 26110 which is very similar to that of pGPL-II of M. habana was also found to terminate with a 3,4-di-O-Me-GlcA.
As is well known for M. avium (Brennan, 1988), the terminal moieties of the oligosaccharides of the polar GPLs are the dominant epitopes recognized by antisera against the whole bacterial cells. Our data show that the anti-M. habana serum includes distinct antibodies reacting specifically with pGPL-I and pGPL-IIϩIII and, among those recognizing the latter, some also cross-reacted with the polar GPLs from M. simiae serotype I strain 26110 and M. avium serovar 19. Thus the differentially O-methylated terminal GlcA residue, or the absence of it, in the pGPLs of M. habana may be sufficiently specific to confer the structural basis upon which it can be distinguished from the closely related M. simiae serotype I and M. avium complex through simple and sensitive serological and biochemical means of identification.
It is not known whether all the other strains identified as M. habana share its distinctive pattern of polar GPLs, but at least two others from Dr. Jenkins' reference collection (referred to as habana-like M. simiae strain 26112 and 26113), were found by us to do so. Since these strains have a different history of culture from TMC 5135 over a period of at least 20 years, it may be concluded that the distinctive pattern is quite stable and that M. habana is distinct from M. simiae serotype I. It seems improbable that the GPLs characterized in the present work are in any way connected with the ability of the strains that produce them to protect against mycobacterial diseases. They are nonetheless important in this context because they apparently make possible a simple, rapid, and unequivocal identification of the proposed vaccine strain, essential for quality control in the eventual production and use of a vaccine. simiae serum (B) with purified pGPLs. Wells of the ELISA plates were coated with 3.5 g (dry mass) per well of purified pGPLs as indicated. The antisera at 1/100 dilution were untreated (Ⅺ), or treated with pGPL-I (o), pGPL-II plus pGPL-III (p), or pGPL-Sim (s) for 2 h at 37°C and then added to the plates at 0.1 ml/well. Bars show mean absorbances at 405 nm (8 replicates) with standard deviations. Absorbance marked with an asterisk are significantly different from absorbances with untreated serum in a 2-sided t test (p Ͻ 0.0001).