Expression of the Core Lipopeptide of the Glycopeptidolipid Surface Antigens in Rough Mutants of Mycobacterium uuium*

Toward studying the genetics, biosynthesis, and roles in the pathogenesis of the dominant surface glycopeptidolipid antigens of Mycobacterium avium, rough colony variants of M. avium serovar 2 were picked, cultured in quantity, and their lipid composi- tion examined. Two of the rough (Rg) variants, Rg-3 and Rg-4, were devoid of glycopeptidolipids or any more elemental structures and thus were similar to those described previously. Two others, Rg-0 and Rg-1, each contained two novel lipopeptides, devoid of any of the carbohydrate substituents that confer serotypic activity on the glycopeptidolipids. The application of gas chromatography, fast atom bombardment-mass spectrometry and ‘H NMR to lipopeptide I established the structure C,,:,-&OH-fatty acyl-D-Phe-D-allo-Thr-D-Ala-L-alaninol. Lipopeptide I1 represented a minor variation of this structure: C,,,,-fi-OH-fatty acyl-D-Phe-D-allo-Thr-D-Ala-L-alaninol. These newly discov- ered lipopeptides are identical to the fatty acyl-tripep-tide-amino alcohol “core” component of the glycopep- tidolipids of the M. avium complex, and thus determine the content of lipids in isolates, lyophilized cells were extracted with CHC13-CH30H (2:l) (20). The total washed lipid fractions (21) were hydrolyzed with 0.1 N NaOH at room temperature to select for alkali-stable lipids (20) and subjected to TLC in mixtures of CHCl, and CH3OH on sheets of precoated Silica Gel 60 (Merck) as described (20). To isolate appre- ciable quantities of lipopeptide for structural characterization, 2.01 g of lyophilized M. auium Rg-1 cells were extracted as described above. The extraction yielded 235 mg of total lipid, and subsequent alkaline hydrolysis of this resulted in 200 mg of alkali-stable lipid. This lipid was dissolved in CHCL and applied to a column (1.5 X 20 cm) of Florisil (60-100 mesh; Sigma) equilibrated in CHCl3. The column was irrigated with 80 ml each of CHCL, followed by 10, 50, and 70% CH3OH in CHCl,. Lipid (30 mg) from the 10% CHSOH fraction was applied to five 20 X 20-cm glass-backed TLC plates (Silica Gel 60 A; 0.25 mm, Whatman) that were developed six times in CHC13-CH30H (12:l). The major lipid bands were observed by spraying with a fine mist of H20. The Silica Gel was scraped from the plates, the lipids were recovered by extraction with CHC13-CH30H (2:1), washed by redissolved and passed Acrodisc PBS containing 0.05% Tween; this murine antibody is specific to M. auium serovar 2 and its particular ssGPL (16). Unattached antibody was removed by three consecutive 5-min washes of PBS (pH 7.2). Antibody specifically bound to the various colony forms was detected by incubation for 1 h in horseradish peroxidase-conjugated goat anti-mouse IgG immu-noglobulin (Cooper Biomedical Inc., Malvern, PA) diluted 1:1,000 in PBS containing 0.05% Tween. The filter paper was washed three times as before. Color development of the labeled colonies was accom- plished by exposure to 4-chloro-1-naphthol and hydrogen peroxide for 10 min. Colonies that did not specifically hind antibody were observed by spraying with ninhydrin and heating at 120 "C.

Expression of the Core Lipopeptide of the Glycopeptidolipid Surface Antigens in Rough Mutants of Mycobacterium uuium* (Received for publication, December 1, 1992, andin revised form, February 2, 1993) John T. Belisle, Michael R. McNeil, Delphi Chatterjee, Julia M. Inamine, and Patrick J. BrennanS From the Department of Microbiology,Colorado State Uniuersity,Fort Collins,Colorado 80523 Toward studying the genetics, biosynthesis, and roles in the pathogenesis of the dominant surface glycopeptidolipid antigens of Mycobacterium avium, rough colony variants of M. avium serovar 2 were picked, cultured in quantity, and their lipid composition examined. Two of the rough (Rg) variants, Rg-3 and Rg-4, were devoid of glycopeptidolipids or any more elemental structures and thus were similar to those described previously. Two others, Rg-0 and Rg-1, each contained two novel lipopeptides, devoid of any of the carbohydrate substituents that confer serotypic activity on the glycopeptidolipids. The application of gas chromatography, fast atom bombardment-mass spectrometry and 'H NMR to lipopeptide I established the structure C,,:,-&OH-fatty acyl-D-Phe-D-allo-Thr-D-Ala-L-alaninol. Lipopeptide I1 represented a minor variation of this structure: C,,,,-fi-OH-fatty acyl-D-Phe-D-allo-Thr-D-Ala-L-alaninol. These newly discovered lipopeptides are identical to the fatty acyl-tripeptide-amino alcohol "core" component of the glycopeptidolipids of the M. avium complex, and thus the Rg-0 and Rg-1 variants represent a form of "deep rough" mutation in M. avium. Separately, we report that these rough variants of M. avium differ genetically from the smooth, virulent form by major deletions of portions of the genes responsible for glycopeptidolipid synthesis (Belisle, J. T., Klaczkiewicz, K., Brennan, P. J., Jacobs, w. R., Jr., and Inamine, J. M. (1993) J. Biol.
Disseminated infections caused by organisms of the Mycobacterium auium complex are among the most common opportunistic infections in patients at advanced stages of human immunodeficiency virus infections (1). Because of convincing evidence that M. auium bacteremia contributes significantly to morbidity and mortality of patients, there is now considerable interest in developing innovative chemoprophylactic strategies for organisms that are extremely difficult to eradicate (2). The cell envelope of members of the M. auium * This work was supported by Grant AI 18357 from the National Institute of Allergy and Infectious Diseases and by Grant AI 30189 under the National Cooperative Drug Discovery Group program from the National Institute of Allergy and Infectious Diseases, National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$To whom correspondence should be addressed. Tel.: 303-491-6700; Fax: 303-491-1815. complex possesses glycolipid antigens, the glycopeptidolipids (GPLs),' which are confined to this set of mycobacteria and are of use in their differentiation from all others (3). The GPLs are located on the surface and are associated with the capsular-like matrix that surrounds the bacillus and forms an electron transparent zone in phagocytic vesicles (3-6). They are thought to suppress lymphocyte blastogenesis (7, 8), a phenomenon observed in M. auium complex infections (9,lO). However, their actual contribution to the pathogenesis of M. auium infections and the intracellular survival of bacilli has been questioned (11, 12), since both the virulent smooth transparent (SmT) and avirulent smooth domed (SmD) colony forms express GPLs, whereas those rough (Rg) variants isolated to date have variously been described as virulent (13) or avirulent (14) and are completely devoid of the surface antigens (15). To resolve obvious confusion and paradoxes, there is a pressing need to develop a range of genetically and phenotypically defined mutants. Such mutants should also lead to an understanding of the pathways for GPL biosynthesis, the underlying genetics responsible for GPL biosynthesis, and the roles of what are copious products in bacterial invasiveness and persistence.
Two forms of glycopeptidolipids are present in the M. auium complex, the ssGPL and nsGPL (3). Both the ss-and nsGPLs have the same basic tripeptide-amino alcohol core comprised of D-Phe-D-allo-Thr-D-Ala-L-alaninol, in which the D-Phe is amide-linked to a fatty acid, and the L-alaninol is glycosidically attached to a 3-0-Me-Rha or a 3,4-di-O-Me-Rha. The primary difference between the nsGPL and ssGPL lies in the extent of glycosylation at the D-allo-Thr constituent. A single 6-deoxytalose unit is attached to the D-allo-Thr in the case of both the nsGPL and ssGPL. In the case of the ssGPL, the 6deoxytalose is further glycosylated to yield a haptenic oligosaccharide that varies in composition among different serovars of the M. auium complex (3). The structural relationship between ns-and ssGPL provides the grounds for the suggestion that the nsGPLs are biosynthetic precursors of the ssGPLs, a hypothesis that is supported by the recent demonstration that a nsGPL of Mycobacterium smegmatis served as a precursor for the synthesis of the recombinant M. auium serovar 2 ssGPL (16). The biosynthesis of the peptidyl core of the GPLs is thought to begin from the fatty acyl unit (17). However, nothing is known of steps beyond this point. This present study describes the isolation and structural characterization of a sugarless lipopeptide core, expressed by some rpugh variants of M. auium serovar 2. Additionally, it is shown that two major classes of spontaneous rough colony variants of M. auium exist, those described previously that are GPL-' The abbreviations used are: GPL(s), glycopeptidolipid(s); ns-, nonspecific; ss-, serovar-specific; Rg, rough; SmD, smooth domed; SmT, smooth transparent; TLC, thin layer chromatography; GC, gas chromatography; MS, mass spectrometry; FAB, fast atom bombardment; PBS, phosphate-buffered saline. negative, lipopeptide-negative, and the present new mutants that are GPL-negative, lipopeptide-positive. The results further demonstrate that the initial step in GPL biosynthesis is the formation of a lipopeptide core, which obviously is the parent of both the ns-and ssGPLs.
A description of the genomic differences that define these two phenotypes is presented separately (18).

EXPERIMENTAL PROCEDURES
Growth of M. auium and Isolation of Morphological Variants-". auium serovar 2 (strain 2151) was originally obtained from the sputum of an individual with pulmonary mycobacterioses. It was initially plated on 7Hll agar (19), and, after 3 weeks of incubation at 37 "C, three morphological variants (SmD, SmT, and a Rg isolate termed Rg-0 (13, 14)) were picked and subcultured on 7 H l l agar. Homogeneous colony morphology was ensured by subculturing each isolate a total of three times. Colonies from the third passage of these strains were scraped, placed in sterile PBS with 0.05% Tween and stirred overnight to generate homogeneous suspensions which were stored as stocks at -70 "C. The rough Rg-1, Rg-3, and Rg-4 isolates, single colonies that did not differ in appearance, were picked from subcultures of the SmD variant plated on 7 H l l agar. They were purified and stored as described above. Large amounts of harvest were obtained by spreading stocks of the SmD, Rg-0, Rg-1, Rg-3, and Rg-4 variants on 7Hll agar plates (15 X 150 mm) at a concentration of 4.0-4.5 X IO3 colony-forming units/plate and incubation at 37 "C for about 3 weeks. The colonial morphology of each culture was examined before harvesting and determined to be consistently smooth domed or rough, as described (13,14). Cells were harvested by scraping, suspended in PBS, autoclaved at 80 "C, and lyophilized.
Purification of Lipopeptides I and ZZ-To determine the content of lipids in isolates, lyophilized cells were extracted with CHC13-CH30H (2:l) (20). The total washed lipid fractions (21) were hydrolyzed with 0.1 N NaOH at room temperature to select for alkali-stable lipids (20) and subjected to TLC in mixtures of CHCl, and CH3OH on sheets of precoated Silica Gel 60 (Merck) as described (20). To isolate appreciable quantities of lipopeptide for structural characterization, 2.01 g of lyophilized M. auium Rg-1 cells were extracted as described above. The extraction yielded 235 mg of total lipid, and subsequent alkaline hydrolysis of this resulted in 200 mg of alkali-stable lipid. This lipid was dissolved in CHCL and applied to a column (1.5 X 20 cm) of Florisil (60-100 mesh; Sigma) equilibrated in CHCl3. The column was irrigated with 80 ml each of CHCL, followed by 10, 50, and 70% CH3OH in CHCl,. Lipid (30 mg) from the 10% CHSOH fraction was applied to five 20 X 20-cm glass-backed TLC plates (Silica Gel 60 A; 0.25 mm, Whatman) that were developed six times in CHC13-CH30H (12:l). The major lipid bands were observed by spraying with a fine mist of H20. The Silica Gel was scraped from the plates, the lipids were recovered by extraction with CHC13-CH30H (2:1), washed by partitioning with water, redissolved in CHCl,-CH,OH (9:1), and passed through an Acrodisc CR, 0.2-+m filter (Gelman Sciences, Ann Arbor, MI) to remove remaining Silica Gel. The ssGPL of smooth variants of M. auium serovar 2 was obtained as described (22).
Analytical Methods-To examine the glycosyl content of lipids, they were hydrolyzed with 2 M CF3COOH and the resulting free sugars converted to alditol acetates and resolved on a DB 23 fused silica capillary column (J&W Scientific, Folsom, CA) as described (23). The amino acid composition of lipids was analyzed by GC of the N,(O)-heptafluorobutyryl isobutyl derivatives as described (24). To establish the enantiomeric form of amino acids, R(-)isobutanol (Aldrich) was used in the derivatization (25). The N,(O)-heptafluorobutyryl isobutyl derivatives were separated on an HP-1 fused silica capillary column (Hewlett-Packard Cetus Instruments).
'H NMR of per-0-acylated lipopeptides was conducted as described (26,27). The molecular weight and amino acid sequence of lipopeptides were determined by FAB-MS. Lipid samples were dissolved in CHCl,-CH,OH (9:l) and 10 pg applied to 3-nitrobenzyl alcohol matrix and examined on a VG 7070 extra high frequency mass spectrometer using an ion saddle field gun operating at 7-8 kV and 1 mA with xenon gas (28,29). Fatty acids were analyzed by first converting them to methyl esters and then to (CH3)3Si ethers (30) prior to analysis by GC-MS on an HP-1 capillary column in a Hewlett Packard 5890A gas chromatogram coupled to a Hewlett Packard model 5970 mass detector.
Colony Dot-Blot-Enzyme-linked Immunosorbent Assay-The procedure was based on that described by Gulig et al. (31) with modifi-cations (32). Colonies were picked from 7H10 agar plates and placed on a sheet of Whatman 40 cellulose filter paper, which was exposed to fumes of 37% formaldehyde for 30 min in a closed container, blocked for 1 h with 2% polyvinylpyrrolidone 40 in PBS (pH 7.2) containing 0.05% Tween, and finally soaked for 4 h in a 1:1,000 dilution of the monoclonal antibody CS-17 in PBS containing 0.05% Tween; this murine antibody is specific to M. auium serovar 2 and its particular ssGPL (16). Unattached antibody was removed by three consecutive 5-min washes of PBS (pH 7.2). Antibody specifically bound to the various colony forms was detected by incubation for 1 h in horseradish peroxidase-conjugated goat anti-mouse IgG immunoglobulin (Cooper Biomedical Inc., Malvern, PA) diluted 1:1,000 in PBS containing 0.05% Tween. The filter paper was washed three times as before. Color development of the labeled colonies was accomplished by exposure to 4-chloro-1-naphthol and hydrogen peroxide for 10 min. Colonies that did not specifically hind antibody were observed by spraying with ninhydrin and heating at 120 "C.

RESULTS
Isolation of Morphological Variants of M. auium-Initial plating of the heavily passaged strain of M. auium yielded primarily SmD colonies. However, the two other colony forms, SmT and Rg, were also present in small numbers. After each colony type had been picked and subcultured to ensure pure colony morphology, colony dot-blot-enzyme-linked immunosorbent assay with the anti-ssGPL-2 monoclonal antibody (CS-17) (16) was applied to each of the colonies. The SmD and SmT variants both reacted strongly to the antibody, demonstrating, as expected, the presence of copious quantities of ssGPLs in each. The initially picked Rg-0 variant did not react with the antibody, again demonstrating, as expected, the absence of GPLs (15). The picked SmD variant, when plated from -70 "C frozen stock at a concentration of 4.1 X IO3 colony-forming units/plate, produced, on average, nine SmT colonies and six Rg colonies per plate. On the other hand, the picked SmT variant, when plated at the same concentration, produced primarily SmT colonies with relatively few Rg colonies (one or two colonies per plate) and only an occasional SmD colony. In contrast, when frozen stocks of the Rg-0 isolate were plated a t a concentration of 4.5 X lo3 colony-forming units/plate, only Rg colonies were observed. Thus, there did not appear to be any reversion of the Rg-0 isolate to smooth morphological forms, either SmT or SmD. Several of the spontaneous Rg colonies arising from the plating of the original SmD variant were picked and subcultured as before. Three of these (Rg-1, Rg-3, and Rg-4) were selected for further study. They demonstrated pure morphology after the third passage, and, like Rg-0, these new rough isolates did not react with the anti-ssGPL-2 monoclonal antibody.
Identification of a New Lipopeptide in Some Rg Variants-The collection of spontaneous Rg variants was examined for differences in lipid content, with a specific focus on what might be truncated forms of the basic GPL structure. Alkalistable lipid fractions from harvests of M . auium SmD and the four rough variants (Rg-0, Rg-1, Rg-3, and Rg-4) were examined by TLC (Fig. 1). Our conjecture that all Rg variants are not identical proved to be correct in that the Rg-0 and Rg-1 variants contained two unique lipids that were absent from Rg-3 and Rg-4. Although the mobility of these two new lipids was similar to that of the nsGPL, the color produced by them in response to the spray was a yellowish brown compared with the bright golden yellow color reflective of the 6-deoxyhexoses within the GPLs (33). This simple observation suggested that the new lipids in Rg-0 and Rg-1 were devoid of the usual sugar components of the GPLs.
The total population of alkali-stable lipids from the four Rg variants was hydrolyzed under conditions appropriate to sugar and amino acid cleavage, derivatized, and examined by nsGPLs-2 -r L SSGPL-2 -II e r n e . .  lanes 4 and 5 ) and also evident in the other variants is a mixture of free fatty acids.
GC; the products from the SmD variant were included for comparison (Fig. 2). The lipids from Rg-0, Rg-1, Rg-3, and Rg-4 lacked the sugars 2,3-di-O-Me-fucose, rhamnose, 6-deoxytalose, and 3,4-di-O-Me-rhamnose, which are present in the combined GPL peptides of the smooth variants of M. avium serovar 2. On the other hand, alkaline-stable lipids from the Rg-0 and Rg-1 variants did contain the characteristic amino acids and amino alcohol of the ssand nsGPLs, whereas those from Rg-3 and Rg-4 were devoid of these amino compounds. These results clearly demonstrated that two distinct types of spontaneous Rg mutants of M. avium had been isolated. One, represented by Rg-3 and Rg-4, was devoid of any elements of the GPLs and corresponded to those described previously (15), whereas the other, typified by Rg-0 and Rg-1, represented a new class of rough M. avium variants and obviously expressed vestiges of the GPL structure in the form, apparently, of the nonglycosylated lipopeptide core.

Chemical Characterization of the Lipopeptides of M. avium
Rg-I-To isolate the putative natural lipopeptide core, total alkali-stable lipid from M. avium Rg-1 was applied to a column of Florisil and eluted with increasing concentrations of CHnOH in CHCln. Examination of the eluates by TLC and by GC for amino acid/amino alcohol composition demonstrated that the lipopeptides appeared mostly in the 10% CHsOH in CHCln eluate. Preparative TLC of this fraction allowed recovery of sizable quantities of two new lipopeptides, lipopeptide I (4.2 mg) and lipopeptide I1 (5.7 mg) (Fig. 3).

FIG. 2. Examination of the glycosyl and amino acid content of the alkaline-stable lipid fraction from morphological variants of M. avium serovar 2 strain 2151. Panel A, GC of alditol
acetates. The temperature program involved an increase from 80 to 160 "C at 30 "C/min followed by 2 "C/min to 180 "C and then 8 "C/ min to 240 "C, which was held for 10 min. IS, an internal standard of 6-0-Me-galactitol acetate. 6d-Tal, 6-deoxytalose. Panel B, GC of N,(O)-heptafluorobutyryl isobutyl esters. The temperature program involved 85 "C for 2 min, an increase a t 8 "C/min to 280 "C which was then held for 8 min. IS, internal standard of derivatized cu-amino adipic acid. The asterisk represents a product of the derivatization procedure and could not be identified by MS. were assigned to the four amide groups of the tripeptideamino alcohol core (Fig. 4A). The exact assignment of these protons to specific amino acids was accomplished by twodimensional COSY 'H NMR (Fig. 4B). The coupling of the amide protons to their respective a-C protons was defined through their connectivity to the protons of the side groups (Table I).
'H NMR also revealed that the fatty acyl function of the lipopeptide was monounsaturated, as seen by the triplet a t 6 5.32 ppm (Fig. 4). The two-dimensional COSY 'H NMR demonstrated the presence of a 0-OH-fatty acid, with the resonance of the 0-C proton apparent at 6 5.01 ppm. This proton was coupled to the a-C protons of the fatty acid which resonated at 6 2.46 ppm and also to the adjacent protons of the fatty acyl chain a t 6 1.45 ppm (Fig. 4). Similar results were obtained with lipopeptide I; however, resolution was not  PCTjOAc, CH,' "The multiplet centered at 6 7.26 ppm can be assigned by its Indicates the two CH, groups of the fatty acid chain which are Indicates the CH, groups of the fatty acid chain. Indicates the single CH, group of the fatty acid chain linked to chemical shift to the aromatic ring protons of Phe (34).
linked to the CH=CH. the a-carbon. native lipopeptides I and I1 were analyzed by FAB-MS (Fig.   5). The native lipopeptide I1 demonstrated a pseudomolecular ion (M + H)' of m/z 871 (Fig. 5B), suggesting that the tripeptide-amino alcohol core was linked to a Cs2-monounsaturated-0-OH fatty acid. The presence of such a fatty acid was consistent with the appearance of the a, m/z 596, bl m/z 624, and b, m/z 725 fragment ions. The a, and b, ions further demonstrated that the D-Phe was N-linked to the hydroxy C32:1 fatty acid, and the bp ion confirmed that the next amino acyl function was D-allo-Thr. The y3+2 m/z 248 and x2 m/z 173 ions demonstrate that the D-allo-Thr carboxyl terminus is linked to the Ala-alaninol terminus; the m/z 230 ion corresponds to the y3 fragment ion minus H,O (Fig. 5, A and B ) .
Lipopeptide I yielded the same y3+2 and xz ions as lipopeptide I1 (Fig. 5A); however, the a', bl, and bS fragment ions and the pseudomolecular ions were all 2 mass units lower, suggesting the presence of a C32-diunsaturated-@-OH fatty acid. enoate) (Fig. 6B). Although positive ion FAB-MS did not produce a distinguishable pseudomolecular ion, a number of fragment ions did confirm that the ssGPL-2 possesses a C32diunsaturated-P-OH fatty acid. The fragment ion of m / z 869 ( Fig. 6A) was identical to the pseudomolecular (M + H)' ion of lipopeptide I and was probably produced by the loss, coupled with hydrogen transfer, of both the 3,4-di-O-Merhamnose and the oligosaccharide component from the serovar 2-specific GPL. Additional loss of the alaninol and H20 probably accounts for the m / z 776 ion. Also, the a,, b,, and b, fragment ions of the ssGPL-2 (Fig. 6A) were the same as those obtained from lipopeptide I.
The fatty acid esters of lipopeptide I1 were obtained by methanolysis. GC-MS analysis of the (CH3)3Si derivative of the methyl esters confirmed that the major fatty acid of lipopeptide I1 was, indeed, a CS2 monounsaturated 3-OH fatty acid (Fig. 7). In addition, this type of analysis revealed that fatty acylation of the lipoprotein I1 is heterogenous in that CB4:,-, C35:1-, and C3&.-3-0H fatty acids were also detected in minor amounts. This heterogeneity was also observed in the FAB-MS of the lipopeptide 11, in which ions 14 and 28 mass units higher than the (M + H)+ ion were observed (Fig. 5).

DISCUSSION
The isolation of spontaneous bacterial mutants on the basis of colony morphology has played a pivotal role in the eluci-dation of biosynthetic pathways for cell wall products and their implication as virulence factors. This point is best exemplified by the deep rough and semirough mutants of Salmonella spp. (35,36), which allowed for the identification of genes and their enzymes responsible for lipopolysaccharide elaboration (35)(36)(37)(38)(39) and the definition of lipid A as the fundamental and toxic core of lipopolysaccharide (40,41). Members of the M. auium complex are well known for their ability to express both rough and smooth colony forms (13,14). However, in comparison with the Enterobacteriaceae, previous biochemical analyses of these morphological variants has not been as fruitful. Initial observations by Fregnan et al. (42) and Schaefer et al. (13) indicated that rough colonyforming variants of M. auium were devoid of ill defined surface structures. A subsequent study by Barrow and Brennan (15), employing rough and smooth variants of M. intracellulare serovar 20, firmly established that a consequence of rough colony formation was the absence of both the ns-and ssGPLs. Barrow (43) subsequently reported the presence of phenylalanine-containing lipopeptides in rough variants arising from cultures of M . auium serovars 4, 8, and 20. However, further analysis revealed the presence of phenylalanine, isoleucine, and alanine and the apparent absence of threonine and alaninol, and thus these products appeared to be unrelated to the GPLs (48).
The present recognition of two nonglycosylated lipopep- The depicted structure of ssGPL-2 has a molecular weight of 1,508. tides, obvious core or elemental forms of the GPL antigens, is the culmination of a search begun several years ago for mutants defective in features of the GPL molecule, particularly glycosyl appendages. Based on the lipopolysaccharide paradigm, we had speculated that formation of a lipopeptide core was the first phase in the biosynthesis of the GPLs (44, 45). However, a thorough search for either water-or lipidsoluble precursors had proved negative. In retrospect, it is now obvious that we had not examined sufficient numbers of the spontaneous rough variants of M. avium-M. intracellulare.
The mechanism of the synthesis of such lipopeptides is unknown. Two separate means exist in prokaryotes for nonribosomal peptide biosynthesis, the principles of either of which may apply to the short tripeptide-amino alcohol core of the GPLs. The first possibility involves a form of direct synthesis in which the amino acids are added directly to an acceptor through the intervention of ATP. The muramyl-tetrapeptide unit of peptidoglycan is synthesized in this way (46). In the case of the peptide antibiotics, the amino acids are attached through thiol groups to a polyenzyme complex, and the peptide bond is subsequently formed through successive pantetheine-aided transpeptidation-transthiolation steps (47). If this latter mechanism were to apply to GPL biosynthesis, then it is likely that the final step in lipopeptide synthesis would involve transfer of the full peptide unit to the fatty acyl function. David et al. (17) demonstrated that the addition of m-fluoro-phenylalanine to cultures of M. avium inhibited GPL biosynthesis and, specifically, the incorporation of radiolabeled amino acids into lipid. They also showed that Dcycloserine inhibited L-Ala racemization to D-Ala, resulting in a 20% inhibition of Ala incorporation into lipid without affecting the incorporation of allo-Thr or Phe. Even though these results do favor the direct synthesis route, additional lipopeptide intermediates of this pathway need to be isolated for definitive proof. From earlier work on the genes encoding the GPL antigens of M. auium (16), it was obvious that the nsGPLs are intermediates of the ssGPLs. However, further work is required to determine whether the oligosaccharide hapten of the ssGPLs is formed on a lipid carrier and then transferred to the nsGPL or whether the sugars are added singly and sequentially to the growing 0-linked oligosaccharide chain of the GPLs.
Clearly, the isolation and characterization of a variety of "deep" rough mutants of M . auium are a major development in the goal of elucidating the biosynthetic pathway of the GPLs. They also provide the means to examine the roles of GPLs in the disease processes induced by M. avium, specifically in eliciting an immunosuppressive response (8).