Biosynthesis of the Linkage Region of the Mycobacterial Cell Wall*

The “core” structure of the cell wall of Mycobacterium and related genera is unique among prokaryotes, consisting of a covalently linked complex of mycolic acids, D -arabinan and D -galactan (mycolylarabinogalactan, mAG), which, in turn, is linked to peptidoglycan via a special linkage unit, - (cid:97) - L -Rha p (1 3 3)- D -GlcNAc-P-. Little is known of the biosynthesis of this complex, although it is the site of action of several common anti-tuberculosis drugs. Isolated cell membranes of Mycobacterium smegmatis catalyzed the incorporation of [ 14 C]GlcNAc from UDP-[ 14 C]GlcNAc into two glycolipids (1 and 2) and of [ 14 C]Rha from TDP-[ 14 C]Rha into glycolipid 2. These products were characterized as polyprenol-P-P-GlcNAc (glycolipid 1) and polyprenol-P-P-GlcNAc-Rha (glycolip-id 2) based on sensitivity of synthesis to tunicamycin, chromatographic characterization of the products of mild acid hydrolysis, and mass spectral analysis of the glycosyl and polyprenyl units. Glycolipids 1 and 2 were shown to be precursors of the linkage unit in polymer- ized cell wall. The inclusion in the assays of UDP-[ 14 C]Gal p and a preparation of cell walls allowed the incorporation of [ 14 C]Gal into two further glycolipids (3

The core or skeletal cell wall of members of the Mycobacterium genus consists of extensively cross-linked peptidoglycan to which is attached the linear D-galactan composed of alternative 5-and 6-linked ␣-D-Galf units (1). Attached in turn to the D-galactan are extensively branched chains of D-Araf-containing arabinan, the distal ends of which are almost completely esterified with mycolic acids (2). We have described this vast macromolecular structure as the mycolylarabinogalactan (mAGP) 1 complex (3). The biogenesis of the mAG portion of this complex is under current investigation because several of the widely used anti-tuberculosis drugs affect aspects of its synthe-sis and resistance to these drugs is a serious world wide public health problem (4). For instance, a target for isoniazid (INH) is a NADH-requiring 2-trans-enoyl fatty acyl reductase apparently on the pathway to mycolic acid biosynthesis and some of the resistance to isoniazid is due to a point mutation in the inhA gene and/or overexpression of the target (5). Ethambutol specifically inhibits the synthesis of the arabinan of AG and of lipoarabinomannan, apparently through its action on a family of arabinosyltransferases (6).
Some clues about the initiation of mAG biosynthesis have arisen from earlier structural work. It has long been known that the AG heteropolysaccharide chains are attached through phosphodiester linkages to C-6 of a proportion of the muramic acid residues of mycobacterial cell walls (7). More recently, chemical analysis of degradation fragments arising from the reducing end of AG obtained from the cell walls of Mycobacterium tuberculosis, Mycobacterium bovis BCG, and Mycobacterium leprae demonstrated the existence of the terminal sequence 35)-D-Galf-(136)-D-Galf-(135)-D-Galf-(134)-L-Rhap-(133)-D-GlcNAc (1,8). Based on the acid lability of the 3-linked GlcNAc unit, the presence of about equal amounts of L-Rhap-(133)-D-GlcNAc and muramyl-6-P in an isolated cell wall fragment and 31 P NMR analysis, it was concluded that the terminal GlcNAc residue is in phosphoryl linkage to the 6-position of some of the muramyl residues of mycobacterial peptidoglycan (8,9). Thus, this aspect of mycobacterial cell wall structure, and, presumably, biosynthesis, shares similarity with the teichoic acid-peptidoglycan complex of many Gram-positive bacteria (10). In view of the role of the mycobacterial linkage unit as the fulcrum of cell wall integrity and as a potential singular site for target-directed chemotherapy against tuberculosis, we set about elucidating its biosynthesis in the belief that such information will give rise to assays amenable to high-throughput screening for new growth inhibitors of M. tuberculosis.

EXPERIMENTAL PROCEDURES
Mycobacteria-Mycobacterium smegmatis mc 2 155 (a gift from Dr. W. R. Jacobs) was grown in a glycerol/alanine salts medium (11). Cells were grown to mid-log phase (about 24 h), harvested, washed with physiologically buffered saline, and stored at Ϫ70°C until required.
Particulate Enzyme Preparations-M. smegmatis (10 g wet weight) was washed and resuspended in a buffer containing 50 mM MOPS (adjusted to pH 8.0 with KOH), 5 mM 2-mercaptoethanol, and 10 mM MgCl 2 (buffer A) (30 ml), at 4°C and subjected to probe sonication (Soniprep 150; MSE Ltd., Crawley, Sussex, United Kingdom; 1-cm probe) at 4°C for a total time of 10 min in 10 60-s pulses with 90-s cooling intervals between pulses. The whole sonicate was centrifuged at 27,000 ϫ g for 12 min at 4°C. The cell wall-containing pellet was resuspended in buffer A to a final volume of 20 ml and divided between two centrifuge tubes. Percoll (Pharmacia, Sweden) was added to each tube to achieve a 60% suspension, and the mixture was centrifuged at 27,000 ϫ g for 60 min at 4°C (12). The particulate, upper, diffuse, cell wall-containing band was collected and washed three times in buffer A and resuspended in this buffer (5 ml) to give the Percoll-60, enzymatically active (12) cell wall fraction, which had a protein concentration of 8 -10 mg/ml.
Membranes of M. smegmatis were obtained by centrifugation of the 27,000 ϫ g supernatant (from the previous step) at 100,000 ϫ g for 1 h at 4°C. The supernatant was carefully removed, and the yellow-pigmented opalescent membranes were gently and superficially washed with buffer A and finally suspended in 0.4 -0.5 ml of this buffer (protein concentration 15-20 mg/ml). Enzymatic Synthesis of dTDP-Rha and dTDP-[U-14 C]Rha-The enzymatic synthesis of dTDP-Rha/dTDP-[ 14 C]Rha was adapted from Okazaki et al. (13). Escherichia coli B ATCC 23848 was grown to mid-log phase as described and harvested after 2 h of growth, washed twice in cold 20 mM Tris-HCl buffer, pH 7.7, at 4°C, disrupted by sonication as described above for M. smegmatis, centrifuged at 27,000 ϫ g, and the supernatant used as a source of enzymes. The reaction mixture contained 2 mM dTDP-Glc (sodium salt; Sigma), 20 mM Tris-HCl (pH 7.7), 0.5 mM Na 2 EDTA, 5 mM MgCl 2 , 10 mM NADPH, and 250 l of the soluble E. coli enzyme extract (13) in a total volume of 500 l which was incubated at 37°C while monitoring NADPH utilization at 340 nm. For the conversion of dTDP-[U-14 C]Glc (237 mCi/mmol; ICN Biomedicals; no longer commercially available) to dTDP-[U-14 C]Rha, 6.15-Ci aliquots of the dTDP-[ 14 C]Rha was added to the above reaction mixture. After 2 h, the conversions were complete and the reactions were quenched by the addition of ethanol (500 l), chilled at Ϫ20°C, and the precipitated proteins removed by centrifugation. The conversion of dTDP-Glc to dTDP-Rha was also followed by chromatography of samples of the supernatant on aluminum backed sheets of Silica Gel (60 F 254 ; E. Merck, Darmstadt, Germany) in 1-butanol/pyridine/acetic acid/ water (5:5:1:3) followed by charring with 10% H 2 SO 4 . Samples were also hydrolyzed with 2 M CF 3 COOH for 2 h at 120°C, alditol acetates prepared and analyzed by GC and GC-MS on a Durabond fused silica column as described (14)  Ϫ form) which was washed with water followed by 50, 150, and 250 mM triethyl ammonium bicarbonate. The dTDP-Rha/dTDP-[ 14 C]Rha appeared largely in the 250 mM fraction. Buffer was removed on a rotary evaporator by repeated evaporation with water followed by CH 3 OH. Final purification of dTDP-Rha/dTDP-[ 14 C]Rha was accomplished by preparative HPLC on a Magnum 9 column (9 ϫ 250 mm) of Partisil 10 SAX (Whatman, Clifton, NJ) at a flow rate of 3 ml/min. The purified product was desalted by gel filtration on a column (1 ϫ 115 cm) of Sephadex G-10 in water, lyophilized, aliquoted in 20% ethanol, and stored at Ϫ20°C. Radiochemical purity was estimated at 96% by analytical HPLC on a Partisil 10 SAX column (250 ϫ 4.6 mm) with a linear gradient of monobasic ammonium phosphate (2- Percoll-60 cell wall enzyme fraction (160 l; 1.3-1.6 mg of protein) and cold dTDP-Rha, UDP-Gal, and UDP-GlcNAc (both from Sigma) were added (final concentrations as described in the text). Other variations to these reaction mixtures are described in the text. To study the effect of tunicamycin (Sigma), the antibiotic was prepared as a sonicated suspension in buffer A (3.2 mg/ml), and 5 l were added to the reaction mixture to achieve a final concentration of 50 g/ml. Reactions were incubated at 37°C for 1 h, followed by the addition of 6 ml of CHCl 3 / CH 3 OH (2:1). Reactions were shaken at room temperature for 20 min, followed by the addition of 680 l of water and centrifuged. The lower organic phase was backwashed with CHCl 3 /CH 3 OH/H 2 O (1:47:48) be-fore application to TLC plates. Distilled and Al 2 O 3 -treated CHCl 3 (to remove traces of HCl in commercial CHCl 3 ) was used throughout.
For purposes of estimating incorporation of radioactivity into polymer, the approach described by McArthur et al. (15) was used. The entire reaction mixture, or a portion of it, was applied to Whatman 3MM chromatography paper which was developed in a descending fashion in isobutyric acid, 0.5 M aqueous NH 4 OH (5:3). The polymer remained at the origin and was counted. The unreacted nucleotide sugar, degraded sugar phosphate, and glycolipid intermediates migrated down the paper (15).
Other Preparatory and Analytical Procedures-The preparation of the mAGP complex from M. bovis BCG cell walls has been described (1,16). The isolation of the linker disaccharide, L-Rha[p(133)-D-GlcNAc, from mAGP has also been described (3). The electron impact-mass spectrum (EI-MS) of the disaccharide showed characteristic A-series fragments (17) Alkali treatment of the 14 C-labeled glycolipids served to demonstrate their alkaline stability and to provide a first step in their purification. Excellent recovery (about 85%) was obtained by dissolving the lipid fraction from reaction mixtures in 0.1 ml of CHCl 3 /CH 3 OH (2:1) followed by the addition of 0.1 ml of 0.2 M NaOH in CH 3 OH and incubation at 37°C for 20 min. Mixtures were neutralized with 2.5 l of glacial CH 3 COOH, dried, suspended in 1.5 ml of CHCl 3 /CH 3 OH (2:1), and 0.25 ml of H 2 O, centrifuged, and the lower (CHCl 3 ) phase retained. First steps in the purification of the glycolipids involved application of the alkali-stable lipids to a column (7 ϫ 0.5 cm) of DEAE-cellulose (acetate form) poured in CH 3 OH and equilibrated in CHCl 3 /CH 3 OH (2:1). The lipid fraction was applied in CHCl 3 /CH 3 OH (2:1) and the column developed with 3 column volumes each of CHCl 3 /CH 3 OH (2:1), CH 3 OH, and 50, 100, 200 mM, and 1 M ammonium formate in CH 3 OH. Salt was removed through biphasic washings (18). TLC of glycolipids was conducted on plates of silica gel in CHCl 3 /CH 3 OH/NH 4 OH/H 2 O (65:25:0.5: 3.6) which were exposed to Kodak X-Omat AR film at Ϫ70°C and subsequently sprayed for the presence of phosphorus with a molybdenum reagent (19) or for polyprenols with a reagent containing p-anisaldehyde (20). Mild acid hydrolysis of glycolipids (21) was conducted on radiolabeled preparations in 100 mM HCl in CHCl 3 /CH 3 OH (2:1) (1 ml) at 20°C for 2 h or 10 mM HCl at 100°C for 10 min. The HCl was neutralized with 0.5 N NaOH before adding H 2 O (150 l) to form a biphase, each phase of which was counted.
Hydrolysis of glycolipids and polymer for neutral sugar content was conducted in 2 M CF 3 COOH at 120°C for 2 h. Hydrolysis of glycolipid preparations for amino sugar content was conducted in 4 N HCl at 100°C for 4 h. Sugars were analyzed in a variety of ways. Radioactive sugar preparations were generally applied to Baker-Flex cellulose plastic sheets (J. T. Baker, Philipsburg, NJ) or sheets from Eastman Kodak and developed three times in formic acid/water/t-butanol/methylethyl ketone (15:15:40:30, by volume) for neutral sugar analysis, and 1-butanol/pyridine/0.1 N HCl (5:3:2) for amino sugar analysis followed by autoradiography. Standards of a variety of sugars (ribose, Ara, Man, Glc, Gal, Rha, GlcNH 2 , GalNH 2 , and mannosamine) were also run on plates and visualized by spraying with phthalic acid, 1-butanol, aniline (22) and heating at 120°C for 5 min. The radioactive monosaccharides released by acid treatment were also assayed by HPLC (Dionex, Sunnyvale, CA) using a gradient pump with pulsed amperometric detection, a Dionex CarboPac PA1 column (4 ϫ 250 mm), and 15 mM NaOH eluent. Fractions were counted, collected, and retention time compared to standards.
Other chromatographic systems have been described (14). However, for analysis of radioactive products (alditol acetates; partially permethylated oligosaccharides) a Durabond (DB)-1 fused silica column (J&W Scientific, Rancho Cordova, CA) was used as described (14,23) but as part of the Hewlett-Packard 5890 Series II Plus Gas Chromatography, coupled to the Lablogic GC-RAM radioactive counter (INUS Systems, Tampa, FL).

Recognition of Two Novel [ 14 C]GlcNAc-containing Glycolipid
Intermediates-The basic assay mixture containing UDP-[U-14 C]GlcNAc described under "Experimental Procedures" was scaled up 4-fold and the reaction mixture subjected to a biphasic organic extraction. About 5% of the radioactivity was incor-porated into the lipid fraction under these conditions. The omission of ATP from the assay had no appreciable effect on incorporation. Addition of decaprenol-P did not increase incorporation. Addition of various non-ionic detergents, such as n-octylglucopyranoside or higher concentrations of ATP, substantially inhibited incorporation. Other major changes in the basic reaction and the consequences are described below.
Samples of the radioactive lipids were applied to silica gel TLC plates which were developed in CHCl 3 /CH 3 OH/NH 4 OH/ H 2 O (65:25:0.5:3.6) and autoradiograms obtained (Fig. 1A). Surprisingly clean products were obtained consisting of two closely migrating glycolipids (GL 1 and GL 2). Conditions that resulted in partial inhibition of [ 14 C]GlcNAc incorporation into the lipid fraction (i.e. prolonged storage of membranes, some detergents) consistently resulted in a more marked inhibition of synthesis of GL 2 (Fig. 1A), suggesting that the formation of GL 2 involved an additional enzymatic step beyond GL 1. Treatment of the GL 1/GL 2 mixture with 0.1 N HCl in CHCl 3 / CH 3 OH (2:1) at 20°C (21) resulted in over 50% loss of lipid radioactivity after 5 min; and ϳ80% loss after 20 min. Treatment of the glycolipids with 0.1 M NaOH at 37°C resulted in 94 and 95% recovery of radioactivity in two experiments, supporting the evidence that these products were polyprenyl-P based glycolipids (24 -27). To confirm that the doublet was glycolipid in nature rather than residual nucleotide sugar or degraded sugar-P, synthesis of GL 1/GL 2 was shown to increase over time (Fig. 1B), and when these products were eluted from the gel, the lipid solubility was confirmed in the two-phase system. Attempts to better resolve the doublet were unsuccessful; most solvents gave the false impression that the product was homogeneous. Thus the evidence pointed to the synthesis of two novel polyprenol-containing glycolipids, much more polar than those described previously in mycobacteria, all of which contained neutral sugars and decaprenol.
Effect of Tunicamycin on Synthesis of the Glycolipid Intermediates-The antibiotic tunicamycin inhibits the transfer of GlcNAc-1-P from UDP-GlcNAc to polyprenyl monophosphates catalyzed by membrane preparations from a variety of organisms including Gram-positive bacteria (28 -30). Thus, in the present instance, inhibition by tunicamycin would implicate a polyprenyl-P-P rather than a polyprenyl-P linkage. Tunicamycin was added to the standard reaction mixture as described under "Experimental Procedures" at a concentration of 50 g/ ml. It had a dramatic inhibitory effect on [ 14 C]GlcNAc incorporation into GL 1/GL 2 ( Table I Table I (3000 cpm) was applied to thin layer plates of silica gel (aluminum backed; 60 F 254 ; E. Merck) and developed in CHCl 3 /CH 3 OH/NH 4 OH/H 2 O (65:25:0.5:3.6). Lane 1, lipid fraction from standard reaction mixture. Lane 2, lipid fraction from standard reaction mixture using membranes partially inactivated by prolonged storage. Autoradiograms were obtained after 5 days of exposure. B, the standard reaction mixture (Table I) Table I, (Table I). However, it was clear from autoradiography that the incorporation of [ 14 C]Rha took place only into GL 2 of the glycolipid doublet (Fig. 2). Incorporation of [ 14 C]Rha into the apolar glycopeptidolipids (GPLs) was also evident (Fig. 2). The GPLs are composed of a lipopeptide containing among others, L-alaninol and D-allothreonine, which provide attachment points for a variety of glycosyl units, among them, invariably, O-methylrhamnosides (32).  It was obvious throughout that the quantities of GL 1 and 2 generated in scale-ups of the standard reaction mixtures were too small for adequate chemical characterization. Accordingly, the standard reaction mixture was increased 158 times (that containing UDP-[ 14 C]GlcNAc was increased 46 times; the corresponding UDP-GlcNAc-cold reaction was increased 112 times). A total of 1.142 ϫ 10 6 cpm were incorporated into lipids. These were applied to three preparative (20 ϫ 10 cm) plates, developed in CHCl 3 /CH 3 OH/NH 4 OH/H 2 O and subjected to autoradiography. The other known mycobacterial sugar-containing polyprenyls, C 35 -P-Man and C 50 -P-Man (24 -26) and C 50 -P-Araf (27), ran much faster in this solvent (R F 0.5-0.7), and hence we knew that the GL 1/GL 2 mixture was not contaminated with other known polyprenyl-containing glycolipids. In excising GL 1 and 2 from the preparative plate, every effort was made to separate them. However, subsequent analytical TLC showed that they were heavily contaminated with each other. The purified products were subjected to mild acid hydrolysis (10 mM HCl, 100°C, 10 min) partitioned within a mixture of CHCl 3 /CH 3 OH/H 2 O (4:2:1), and the lipid in the organic phase subjected to MS analysis as described (33). EI-MS of the product from the GL 1-rich preparation showed an apparent molecular ion peak at m/z ϭ 484 (M-H 3 PO 4 -sugars) ϩ , similar to that of the octahydroheptaprenol fragment arising from the 6-Omycolyl-␤-D-mannopyranosyl-1-monophosphoryl-octahydroheptaprenol (Myc-PL) recently described (33). A dominant ES-MS fragment (m/z 582) in the GL 2-rich preparation was also indicative of octahydroheptaprenol-P [(M-sugar(s) ϩ (H)] ϩ (27). Hence both products seemed to contain the C 35 octahydroheptaprenol rather than the decaprenol usually found in mycobacterial products (24 -27, 34), or undecaprenol. The proposed structures, C 35 -P-P-GlcNAc and C 35 -P-P-GlcNAc-Rha, would explain the much greater polarity of GL 1/GL 2 compared to C 50 -P-Manp, C 35 -P-Manp, or C 50 -P-Araf. However, much greater quantities of the two glycolipids are required for more thorough characterization of the type recently applied to the polyprenyl of the C 50 -P-Araf (27)  . Total volume, 160 l. Ten such tubes were installed, incubated at 37°C for the indicated times, stopped by the addition of 1 ml of C 2 H 5 OH, and the entire reaction mixture applied to sheets of Whatman 3MM paper which were developed overnight in isobutyric acid/NH 4 OH (5:3), cut into strips and counted. The origin represented the polymer, while the solvent front contained the glycolipids (8). a A mixture of GL 1 and GL 2 (the 40 -100 mM ammonium formate in methanol fraction from DEAE cellulose) was dissolved in 2% Nonidet P-40 (35 l) and 15 l of this mixture (42,800 cpm) was added to the reaction mixture (containing 100 l of membrane preparation (1.5 mg of protein); 143 l of the P-60 cell wall preparation (1.1 mg of protein); 0.1 mM UDP-Gal and buffer A to a final volume of 320 l; 0.1 mM UDP-GlcNAc and 0.05 mM TDP-Rha were included, although apparently not required). Incubation was for 1 h at 37°C. The whole reaction mixture was applied on Whatman 3MM paper, which was developed in isobutyric acid; 0.5 M ammonium hydroxide (5:3) overnight. The origins of the paper strips, containing the polymerized cell wall, were cut out and counted. NAc present in GL 1/GL 2 into this polymer was observed (Table II). In order to examine the nature of the polymer synthesized under these conditions, membranes were incubated with UDP-[ 14 C]GlcNAc in the presence of 0.1 mM UDP-Gal and 0.05 mM dTDP-Rha and the Percoll-60 cell wall fraction for various times. The entire reaction mixtures were applied to strips of Whatman 3MM and chromatographed overnight in isobutyric acid/0.5 M NH 4 OH (5:3). The areas around the solvent front, corresponding to GL 2/GL 3 were excised and counted. Likewise, the material at the origin, the cell wall polymer, was counted. Incorporation into both populations was linear over the course of the experiment (Fig. 4). Thus, the kinetics were more reminiscent of the relationship between the dolicholbound oligosaccharide precursors and the core region of yeast mannoproteins (36), which also involves a GlcNAc-containing (chitobiose) linkage (37), than that of the simpler mycobacterial polyprenol-P-Man precursors and mycobacterial mannan (24,34), indicative of a greater similarity to yeast mannoprotein synthesis (37). The labeled polymer was hydrolyzed, subjected to cellulose TLC and autoradiography as described for GL 1/GL 2. Only GlcNH 2 was present; there was no evidence for synthesis of muramic acid and hence of peptidoglycan, under these conditions. Application of the approach (Fig. 3 and Ref. 3) used to identify the Rha(133)-[ 14 C]GlcNAc linkage region in the GL 1/GL 2 mixture produced the radiolabeled disaccharide from the polymer.
Higher Glycolipid Intermediates-The addition of the cell wall enzyme preparation (Percoll-60) to the standard reaction mixture resulted in a slight inhibition of incorporation of [ 14 C]GlcNAc into lipids; certainly there was no enhancement of activity as was expected (Table III). However, the presence of Percoll-60 in the assays had a dramatic qualitative effect on the profile of glycolipids synthesized in that TLC showed the emergence of other new, more polar glycolipids, GL 3 and GL 4 (Fig. 5). These new products were similar to GL 1 and GL 2 in terms of acid lability and alkaline stability, and thus it seemed likely that the higher GL 3 and GL 4 were more glycosylated, specifically galactosylated, versions of GL 1 and GL 2. To further prove the point and identify the nature of the new glycosyl substituents, a series of reactions were installed con-taining membranes, the Percoll-60 cell wall fraction, cold nucleotide sugars, and UDP-[ 14 C]GlcNAc, TDP-[ 14 C]Rha, or UDP-[ 14 C]Gal. The lipids were extracted, treated with alkali, and subjected to TLC and autoradiography (Fig. 6). The effects of the new adducts (Percoll-60 and higher concentrations of all likely nucleotide sugar precursors) to this reaction mixture were decidedly obvious in the UDP-[ 14 C]GlcNAc-containing assay with the clear-cut emergence of the higher glycolipid homologs, GL 3 and GL 4 (Fig. 6, lane 1). These higher homologs were also faintly evident in the [ 14 C]Rha labeling experiment (lane 2).
However, the inclusion of UDP-[ 14 C]Galp in the assay had the most dramatic effect. Only GL 3 and Gl 4 and material at the origin, perhaps higher homologs, became labeled, indicating growth of the galactan chain on the polyprenol-P-P-Glc-NAc-Rha unit. The lipid products from these three reactions were subjected to mild acid hydrolysis and the water soluble products chromatographed on paper against the Rha(133)-GlcNH 2 standard (Fig. 7). The profile bears out the conclusion that [ 14 C]Gal from UDP-[ 14 C]Gal was selectively incorporated into higher glycolipid intermediate(s). Radioactive GC of the NaB[ 2 H] 4 reduced, methylated and acetylated [ 14 C]Gal oligosaccharide preparation, as described for the derivatized linkage disaccharide (Fig. 3), confirmed the dominance of products with retention times indicative of [ 14 C]Gal-containing tri-and tetrasaccharide.
In order to demonstrate that the [ 14  In these experiments, no direct evidence was provided that GL 3 and GL 4 were derived from GL 1/GL 2, i.e. a precursorproduct relationship was not demonstrated. In order to generate preliminary evidence to this effect, tubes containing UDP-[ 14 C]GlcNAc and the standard reaction mixture were incubated at 37°C for 30 min, followed by the addition of "cold" UDP-Gal as a substrate for further synthesis and more UDP-GlcNAc as a chase. Tubes were then incubated further for variable times (Fig. 8). The emergence of the [ 14 C]GlcNAc-containing GL 3 and GL 4 and also, apparently, a GL-5 was evident, particularly after the longer incubation periods. The other most dis-tinctive quantitative effect of this form of chase was a steady loss of radioactivity from the total lipid fraction (175,000 cpm/ reaction mixture/0.8 mg of protein at O chase time (tube 1), compared to 70,000 cpm after the 90 min chase). Over this period, the GL 1/2 combination lost over half of its radioactivity, and incorporation into GL 3 and GL 4 increased 4-fold and 15-fold, respectively. DISCUSSION Present success in defining the early stages of mycobacterial cell wall synthesis arose from the realization of chemical, and hence biosynthetic, similarities with the cell walls of Grampositive bacteria (38). Early chemical studies of Gram-positive cell walls had revealed that teichoic acids released from cell walls by treatment with dilute acid contained a phosphate group esterified at C6 of some of the muramic acid residues in the wall peptidoglycan (7). Subsequent investigation of the ribitol teichoic acid of Staphylococcus aureus demonstrated an attachment to peptidoglycan by a discrete "linkage unit" containing GlcNAc-1-P and 2 or 3 glycerol-P residues (39). From detailed analysis of teichoic acid attachments in a wide range of species, it is now clear that linkage units consist of a disaccharide-1-phosphate (N-acetylmannosaminyl-N-acetylglucosamine-1-phosphate) unit with a small number (1-3 depending on the species) of glycerol-P residues attached to the N-acetylmannosamine (ManNAc) (10). This unit is attached, in turn, to muramic acid in peptidoglycan through the GlcNAc-1-P, while the teichoic acid chain proper, composed of ribitol-P units, is linked through a phosphodiester to the terminal glycerol-P residue of the linkage unit. The structure is thus: (ribitol-P) n -(glycerol-P) 1-3 -ManNAc-GlcNAc-1-P-MurNAc . . . (40). Of more direct relevance to this study, linkage units are also involved in cell wall attachment of a polysaccharide in Micrococcus luteus, and of teichoic acids in actinomycete and related bacteria (10). The GlcNAc-1-P link is highly susceptible to acid hydrolysis, accounting for the ease of extraction of teichoic acid from the cell wall under acidic conditions and the retention of a phosphate group on muramic acid. The similar susceptibility of the mycobacterial arabinogalactan-peptidoglycan linkage, and the isolation of (Galf) 1-3 Rha-GlcNAc units (8) demonstrated the existence of distinct but analogous linkage units in mycobacteria. Thus, current evidence indicates that the whole of the mycolylarabinogalactan (mAG) complex of mycobacterial cell wall is covalently linked to peptidoglycan through a crucial  Fig. 6 were subjected to mild acid hydrolysis (0.1 N HCl in CHCl 3 /CH 3 OH (2:1), 25°C, 4 h) and neutralized (recovery of water-soluble radioactivity was 84%, 27% (due to the presence of the alkali-stable GPLs) and 83%, respectively). Radioactivity was applied to sheets of Whatman 3MM, developed in isobutyric acid/0.5 M NH 4 OH (5:3) overnight, and the strips cut into 1-cm sections and counted. GlcNH 2 and linkage disaccharide (Rha-GlcNH 2 ) were run on parallel strips and located with aniline-phthalate.  6). The lipids were extracted from reaction mixtures, chromatographed, exposed to x-ray film for 2 weeks as described in the legend to Fig. 1, and individual lanes also counted. disaccharide linker unit attached to the nonreducing terminus of the galactan of mAG in the following arrangement: -D-Galf-(␣134)-␣-L-Rhap-(133)-D-GlcNAc-(13 P36)Mur-N-glycolyl (8). Presumably only the occasional muramic acid residue is so occupied (8). Yet to be resolved is the old evidence that attachments not involving phosphorus also exist (41). The special version of linkage unit formed in mycobacteria also extends to a broad range of Mycobacterium, Rhodococcus, and Nocardia spp. (42,43).
The type of linkage unit that attaches teichoic acid to peptidoglycan in the Gram-positive cell wall is highly conserved among a wide range of only distantly related species indicating that it confers a significant advantage in either the synthesis or the properties of the cell wall (10). Likewise, we have attributed comparable significance to the mycobacterial linkage unit (38). The initial studies of teichoic acid synthesis in S. aureus demonstrated the key role played by the linkage unit in initiation of new teichoic acid polymer chains (10). A membrane fraction from mechanically disrupted bacteria catalyzed the synthesis of a trace of ribitol teichoic acid from the precursor CDP-ribitol. However, addition of the precursors of the linkage unit, UDP-GlcNAc, and CDP-glycerol, dramatically stimulated teichoic acid synthesis. More detailed investigations, including pulseradiolabeling experiments, showed that the biosynthetic system catalyzed the incorporation of GlcNAc-1-P from UDP-Glc-NAc into a lipid molecule which in turn gave rise to other lipids in the presence of CDP-glycerol. When CDP-ribitol was added, radioactivity from these lipids appeared in the newly synthesized teichoic acid (44). The lipophilic part of the lipids was shown to be an undecaprenolphosphate of the type involved in peptidoglycan synthesis, and the initial transfer to it of Glc-NAc-1-P was very sensitive to inhibition by the antibiotic tunicamycin (44). In the case of the M. luteus polysaccharide, the initial biosynthetic reaction is also the tunicamycin-sensitive transfer of GlcNAc-1-P to a polyprenyl phosphate carrier lipid (45). Thus, these experiments demonstrated that the first stage in teichoic acid and M. luteus polysaccharide synthesis was the assembly of linkage unit on the polyisoprenol-phosphate carrier lipid, and that this linkage unit-lipid then acted as the primer on which the polymer was assembled from CDP-ribitol.
The evidence presented in this report that these steps represent the initial events in mycobacterial cell wall biogenesis is firm, although based mostly on comparative radiolabeling experiments, and sensitivity of products to acid, base, and tunicamycin. The paucity of tangible quantities of polyprenyl-P-P-GlcNAc (the proposed structure for glycolipid 1) and polyprenyl-P-P-GlcNAc-Rha (glycolipid 2) precluded characterization of the polyprenyl component. The answer to this question may lie in the isolation of glycolipid 2, which appears present in relatively large, steady state levels. The question of the nature of the polyprenyl carrier is an important one, since mycobacteria to date have yielded only decaprenols (24,25,27) and heptaprenols (25,33) as their version of the bactoprenols, but never the common undecaprenol.
Subsequent steps in cell wall biogenesis are less well estab-lished. UDP-Galp is a very effective substrate for the Galfcontaining galactan component of mAG 2 and, according to present results, of the apparent glycolipid intermediates. In Salmonella enterica, UDP-Galp is a precursor of the Galf in T1 polysaccharides (46). In Penicillium charlosii, the Galf of UDP-Galf can be transferred to the polysaccharide galactocarylose (47). Stevenson et al. (48) based on genetic analysis concluded that a single enzyme can convert UDP-Galp to UDP-Galf through a 2-keto intermediate and that orf6 is the gene involved. Present results would indicate that this enzyme is membrane-bound. Recently, we described the presence of the rfb (rhamnose biosynthetic) genes close to a new insertion-like element (ISL445) in the genome of M. tuberculosis, which included the 3Ј-region of the rfbB gene, the whole rfbC gene, and the 5Ј region of the rfbA gene; however, the rest of the rfbA gene and the following rfbD gene were not obvious in this region. 3 Accordingly, the genetics and enzymology of mAG-linkage unit synthesis show the promise of unique molecular principles to match the novelty of the biochemical pathway described herein. Thus, this work represents a return to the topic of mycobacterial cell wall biogenesis which proved an intractable problem in previous times, and the assays and intermediates described pave the way for screens for new anti-tuberculosis drugs to counteract the serious problem of drug resistance.