Biosynthesis of the Galactan Component of the Mycobacterial Cell Wall*

The structural core of the cell walls ofMycobacterium spp. consists of peptidoglycan bound by a linker unit (-α-L-Rhap-(1→3)-D-GlcNAc-P-) to a galactofuran, which in turn is attached to arabinofuran and mycolic acids. The sequence of reactions leading to the biogenesis of this complex starts with the formation of the linker unit on a polyprenyl-P to produce polyprenyl-P-P-GlcNAc-Rha (Mikušová, K., Mikuš, M., Besra, G. S., Hancock, I., and Brennan, P. J. (1996) J. Biol. Chem. 271, 7820–7828). We now establish that formation of the galactofuran takes place on this intermediate with UDP-Galf as the Galf donor presented in the form of UDP-Galpand UDP-Galp mutase (the glf gene product) and is catalyzed by galactofuranosyl transferases, one of which, theMycobacterium tuberculosis H37Rv3808c gene product, has been identified. Evidence is also presented for the growth of the arabinofuran on this polyprenyl-P-P-linker unit-galactan intermediate catalyzed by unidentified arabinosyl transferases, with decaprenyl-P-Araf or 5-P-ribosyl-PP as the Arafdonor. The product of these steps, the lipid-linked-LU-galactan-arabinan has been partially characterized in terms of its heterogeneity, size, and composition. Biosynthesis of the major components of mycobacterial cell walls is proving to be extremely complex. However, partial definition of arabinogalactan synthesis, the site of action of several major anti-tuberculosis drugs, facilitates the present day thrust for new drugs to counteract multiple drug-resistant tuberculosis.

enzymes and endogenous cofactors in M. smegmatis (11). dTDP-Glc (sodium salt; Sigma; 12 nmol) was incubated at 37°C for 1 h with 25 l of the 100,000 ϫ g supernatant of disrupted M. smegmatis (70 g of cytosolic protein), followed by the addition of 100 l of cold ethanol. After 20 min on ice, the sample was centrifuged at 14,000 ϫ g, and the supernatant was removed and evaporated under a stream of N 2 . The dried material was dissolved in 25 l of deionized water and used as a source of dTDP-Rha. Conversion of dTDP-Glc to dTDP-Rha under these conditions was about 95%, as revealed by analytical HPLC on a Partisil 10 SAX column as described (7). dTDP-[ 14 C]Rha was prepared from [U- 14 C]sucrose by conversion of the glucose moiety of the sucrose to [U-14 C]glucose-1-phosphate by sucrose phosphorylase followed by further conversion to dTDP-[ 14 C]Glc by ␣-D-glucose-1-phosphate thymidylyl transferase (RmlA) and to dTDP-[ 14 C]Rha by RmlB-D. The ␣-D-glucose-1-phosphate thymidylyl transferase was prepared from M. tuberculosis rmlA (11)  Then 700 l of absolute ethanol were added, and the precipitated protein was removed by centrifugation at 14,000 ϫ g for 5 min. The bulk of the ethanol was removed by evaporation, and the dTDP-[ 14 C]Rha was purified by HPLC as described (12).
Preparation of Enzymatically Active Membranes and Cell Envelope-M. smegmatis mc 2 155 and M. smegmatis pJJV7-3808c were grown as described (7). Enzymatically active membranes and cell envelope (wall and membrane) were prepared as follows. Cells (10 g) were suspended in 50 mM MOPS buffer, pH 7.9, containing 5 mM 2-mercaptoethanol and 10 mM MgCl 2 (buffer A), subjected to probe sonication (7), and centrifuged at 23,000 ϫ g for 20 min at 4°C. The pellet was resuspended in buffer A, and Percoll (Amersham Pharmacia Biotech) was added to achieve a 60% suspension, which was centrifuged at 23,000 ϫ g for 60 min at 4°C. The white upper band was isolated, and Percoll was removed by repeated suspension in buffer A and centrifugation. The fraction (cell envelope) was resuspended in buffer A to a protein concentration of 8 -10 mg/ml for use. Membranes were obtained by centrifugation of the 23,000 ϫ g supernatant at 100,000 ϫ g for 75 min at 4°C and suspended in buffer A to give a protein concentration of 15-20 mg/ml.
Reaction Mixtures and Fractionation of Reaction Products-The reaction mixtures for assessing [ 14 C]Gal incorporation consisted of UDP-[U-14 C]Galp (PerkinElmer Life Sciences; 325 mCi/mmol, 1 Ci), which was dried under a stream of N 2 , dissolved in 36 l of buffer A, and preincubated with 4 l of the UDP-Galp mutase (0.13 mg of protein) at 37°C for 15 min, followed by 20 M UDP-GlcNAc, 20 M dTDP-Rha, 60 M ATP, membranes (1-2 mg of protein), cell envelope fraction (0.7-1.5 mg of protein), and buffer A to a total volume of 320 l. After incubation of the reaction mixture for 1 h at 37°C, CHCl 3 /CH 3 OH (2:1; 6 ml) was added, which was left rocking at room temperature for 10 min and centrifuged (3000 ϫ g). The CHCl 3 /CH 3 OH phase was removed from the pellet and treated as described below. To remove residual UDP-[ 14 C]Gal from the pellet, CH 3 OH, 0.9% NaCl (1:1; 2 ml) were added, and the mixture was bath-sonicated for 1 min and centrifuged at 3000 ϫ g. The supernatant was discarded, and the pellet was further extracted with 50% CH 3 OH in H 2 O and 100% CH 3 OH, which were also discarded. The washed pellet was extracted with CHCl 3 /CH 3 OH/H 2 O (10:10:3) (13) to remove more polar products (the lipid-linked polymer) and finally with "E-soak" (water/ethanol/diethyl ether/pyridine/concentrated ammonium hydroxide; 15:15:5:1:0.017) (14) to obtain [ 14 C]Gal-labeled lipidlinked products of even greater polarity. The insoluble pellet was suspended and stored in 1 ml of E-soak. To the CHCl 3 /CH 3 OH (2:1) extract, 680 l of deionized water was added to achieve a biphasic mixture. The upper aqueous phase was removed and discarded, and the bottom phase was backwashed with CHCl 3 /CH 3 (16)) to the reaction mixture. Alternatively, the immediate Araf donor, C 50 -P-Araf, was used in a modified reaction mixture, containing C 50 -P-[1-14 C]Araf; prepared as described (17) (300,000 cpm); and redissolved in 15 l of 2% Nonidet P-40, 60 M ATP, 100 M UDP-GlcNAc, 100 M dTDP-Rha, 100 M UDP-Galp (preincubated with the UDP-Galp mutase enzyme; 0.13 mg of protein), 2 mg of the membrane protein, and 1.6 mg of the cell envelope preparation in a total volume of 320 l. The mixture was incubated at 37°C for 2 h and extracted as described above.
In order to examine the precursor role of GL-1/GL-2, GL-1/GL-2 were synthesized in situ (in membranes), by incubating membrane protein (6 mg) with 5 Ci of UDP-[U-14 C]GlcNAc, 20 M dTDP-Rha, 60 M ATP, and buffer A in a total volume of 1.6 ml, for 1 h at 37°C. Two such reaction mixtures were prepared, which were combined in a 40-ml Beckman centrifuge tube, diluted with about 30 ml of buffer A, and centrifuged at 100,000 ϫ g for 75 min at 4°C. The supernatant was discarded, and the membranes containing radiolabeled GL-1/GL-2 were stored at Ϫ20°C prior to use. Membranes were suspended to a final volume of about 600 l and used as a substrate in a time course experiment to follow the conversion of the GL-1/GL-2 precursors to lipid-linked polymer. The assay conditions were as follows. Glass tubes containing 80 l of radiolabeled membranes (100,000 cpm), 60 M ATP, 20 M UDP-Galp (preincubated with 0.13 mg of UDP-Galp mutase), membrane preparation (1.9 mg of protein) and cell envelope preparation (0.8 mg of protein) in a total volume of 320 l in buffer A were incubated for 10, 20, 40, and 80 min. The 0-min time point was established in a reaction mixture containing membranes and cell envelope preparation that had been boiled for 30 min. The incubations were stopped by adding CHCl 3 /CH 3 OH (2:1; 6 ml) to the reaction mixture and extracted as described above.
Analytical Procedures-DEAE-cellulose (acetate) chromatography of the lipid-linked polymer was performed in a Pasteur pipette containing 1 ml of resin equilibrated with 60% CH 3  Mild acid hydrolysis of the CHCl 3 /CH 3 OH/H 2 O (10:10:3)-and Esoak-soluble lipid-linked polymers was conducted on dried samples (ϳ5,000 -30,000 cpm) suspended by bath sonication in 50 l of 1-propanol with 100 l 0.02 N HCl at 60°C for 30 min (18) and then neutralized with 10 l of 0.2 N NaOH. Products released by mild acid treatment of the lipid-linked polymer were analyzed on a column (1 ϫ 118 cm) of BioGel P-100 resin (Bio-Rad) in 0.1 M sodium acetate (pH 7.0).
Complete acid hydrolysis for purposes of [ 14 C]sugar identification was conducted on samples (ϳ1500 cpm) in 2 M CF 3 COOH for 1 h at 120°C. Hydrolysates were analyzed by TLC on silica gel plates (Merck) developed twice in pyridine/ethyl acetate/glacial acetic acid/water (5:5: 1:3). TLC plates were subjected to autoradiography using Kodak BioMax MR film. Cold sugar standards were visualized by charring with 10% cupric sulfate in 8% phosphoric acid.
Methylation of the [ 14 C]Gal-labeled lipid-linked polymer was accomplished by the NaOH method (19). The per-O-methylated [ 14 C]Gallabeled lipid-linked polymer was hydrolyzed in 2 M CF 3 COOH at 110°C for 2 h, the acid was evaporated, and sugars were reduced with NaB SDS-PAGE was performed on 15% polyacrylamide gels or on commercial 10 -20% gradient Tricine SDS-polyacrylamide gels obtained from Novex (San Diego, CA), under conditions recommended by the manufacturer. Blotting to nitrocellulose was performed at 50 V for 1 h.

Requirement for UDP-Galf for Galactan Synthesis-Previ-
ously, we had demonstrated the biosynthesis of polyprenyl-P-P-GlcNAc (GL-1) from endogenous polyprenyl-P and UDP-Gl-cNAc, followed by synthesis of polyprenyl-P-P-GlcNAc-Rha (GL-2) with the addition of TDP-Rha (7). Further polyprenyl-P-P-GlcNAc-Rha-Gal (GL-3) and polyprenyl-P-P-GlcNAc-Rha-Gal-Gal (GL-4) were formed from the newly synthesized polyprenyl-P-P-GlcNAc/polyprenyl-P-P-GlcNAc-Rha (GL-1/GL-2) in the presence of added UDP-Galp, mycobacterial membranes, and cell envelope fraction. In E. coli K12 (9), Klebsiella pneumoniae (20), and M. smegmatis (8), UDP-Galf is formed by a one-step transformation of UDP-Galp catalyzed by UDP-Galp mutase (EC 5.4.99.9), the product of the glf gene. To differentiate the role of membranes from cell envelope in the biosynthesis of GL-1 to -4 and to examine the role of UDP-Galp mutase, experiments were conducted as described in Fig. 1. Thoroughly washed membranes alone produce GL-1 to -4 if UDP-Galp mutase is included in the reaction (Fig. 1, A and B). Membranes in the absence of exogenous UDP-Galp mutase showed only slight incorporation (about 10%) of [ 14 C]Gal from UDP-[ 14 C]Galp into the glycolipid fraction. The results also demonstrate that the presence of dTDP-Rha stimulated the formation of GL-2 and that TDP-Rha, which is not commercially available, can be replaced by dTDP-Glc and cytosol followed by inactivation of the cytosolic enzymes to avoid catabolism of nucleotide sugars. The results confirm the role of cytoplasmic RmlB (dTDP-Glc 4,6-dehydratase), RmlC (dTDP-6-deoxy-4-ketoglucose epimerase), and RmlD (dTDP-Rha synthase) and membrane rhamnosyltransferase in synthesis of the Rha of the LU, in accordance with the presence of the corresponding genes (rmlB to -D and wbbL) in the M. tuberculosis genome (21). The presence of tunicamycin in reaction mixtures drastically inhibited synthesis of the [ 14 C]Gal-labeled GLs (Fig.  1C), confirming the role of polyprenyl-P in this synthesis.

Incorporation of [ 14 C]Gal from UDP-[ 14 C]Galp into Galactofuran-With the realization that the apparent [ 14 C
]Gal-containing intermediates of galactan synthesis were lipid-linked, solvents developed for solubilization of dolichyl-P-P-linked oligosaccharides (13) and phytosphingoglycolipids (14) were applied in search of more polymerized galactan intermediates. After extraction with CHCl 3 /CH 3 OH (2:1) and subsequent washing, the pellet was extracted with CHCl 3 /CH 3 OH/H 2 O (10:10:3), followed by water/ethanol/diethyl, ether/pyridine/ NH 4 OH ("E-soak") ( Table I). About 20% of the applied radioactivity was incorporated into [ 14 C]Gal-labeled products, of which about 8% was in CHCl 3 /CH 3 OH (2:1), 40% in CHCl 3 /CH 3 OH/ H 2 O (10:10:3), 50% in E-soak fractions, and about 2% in the final pellet; the distribution of the counts between the two polar solvents varied from experiment to experiment. A surprising outcome of this approach was the paucity of radioactivity in the insoluble peptidoglycan-bound galactan and the preponderance in lipid-soluble material. TLC of the CHCl 3 /CH 3 OH-soluble products in CHCl 3 , CH 3 OH, NH 4 OH, 1 M ammonium acetate, H 2 O (180:140:9:9:23) showed the presence of a ladder of GLs of increasing polarity, indicative of sequential glycosylation of GL-1/GL-2, apparently 1 Galf unit at a time, but apparently to a finite length of about 4 Galf units. The products all demonstrated mild acid sensitivity and mild alkali resistance (7), in accordance with polyprenyl-P linkage. The more highly glycosylated lipid-linked polymers in the CHCl 3 /CH 3 OH/H 2 O and E-soak solvents did not migrate under these conditions (Fig. 2).
The relative contributions of the cell envelope and membrane fractions to incorporation of [ 14 C]Galf into these extracts were examined (Table I). The absence of the cell envelope enzyme fraction had a profound effect on [ 14 C]Gal incorporation into the more polar, presumably more glycosylated CHCl 3 /CH 3 OH/ H 2 O (10:10:3) and E-soak-soluble products. The omission of membranes from the reaction mixture still allowed significant synthesis of [ 14 C]Gal-labeled polymer, indicating that the cell envelope fraction contained all of the enzymes involved in biosynthesis of these products. Inclusion of tunicamycin, which is known to inhibit transfer of GlcNAc-1-P from UDP-GlcNAc to dolichyl-P/polyprenyl-P (22)(23)(24), dramatically inhibited production of the [ 14 C]Gal-labeled polymer (Table I), confirming that these reactions are polyprenyl-P-based and probably serve as the initiation point for galactan biosynthesis. The relatively smaller inhibition by tunicamycin of incorporation of radioactivity into the CHCl 3 /CH 3 OH (60%) compared with the CHCl 3 / CH 3 OH/H 2 O (94%) and E-soak (88%) extracts was probably due to synthesis of galactolipids (the galactosyl diacylglycerides) other than GL-3 to -5 (see Fig. 1C).

Properties of the Lipid-linked [ 14 C]Gal-labeled Polymer-Gel filtration of the CHCl 3 /CH 3 OH/H 2 O and E-soak extracts on
BioGel P-100 resulted in poor recovery of material in the void volume, properties compatible with lipid micelles (Fig. 3A). Mild acid hydrolysis under conditions suitable for the release of oligosaccharides bound to polyprenyl-P-P (19) resulted in products that were included in the column (Fig. 3B), were of substantial molecular weight, and showed a recovery of about 85%. There was a clear difference in size of the polymer released from the CHCl 3 /CH 3 OH/H 2 O (10:10:3) compared with E-soakextracted material, suggesting that the former is an incompletely glycosylated version of the latter. Since deacylated lipoarabinomannan (ϳ17 kDa) (25) eluted at 43 ml and deacylated lipomannan (ϳ8 kDa) eluted at 64 ml from this M ATP, and buffer A in a total volume of 320 l. Tunicamycin was added to one of the reaction mixtures to a final concentration of 50 g/ml, as described in Table I. Aliquots of 5% of total lipid extracts were chromatographed as above. Lane 1, control (1500 cpm); lane 2, tunicamycin-treated (600 cpm). TLC plate was exposed to film for 8 days. The unmarked products in C are alkali-labile glycosyldiacylglycerols. column, the approximate mass of the larger polymer is ϳ10.8 kDa, and the smaller polymer is about the same mass as the mannan from lipomannan, i.e. ϳ8 kDa. The size of the mature AG released from the mycobacterial cell wall is of the order of 15 kDa (5), and thus the polyprenyl-P-P-linked polymer generated by the in vitro system is apparently not fully glycosylated. The acidic nature of the population of lipid-linked polymer macromolecules was confirmed by chromatography on DEAEcellulose; about 85% were recovered by elution with 50 mM ammonium formate in 60% CH 3 OH in H 2 O with 0.1% NH 4 OH.
Despite their sizes, the lipid-linked polymers migrated on a silica gel TLC plate in 60% CH 3 OH in H 2 O containing 0.025% NH 4 OH (Fig. 4A). The oligosaccharides released by mild acid hydrolysis were also visualized by this means. SDS-PAGE of the lipid-linked polymers demonstrated the heterogeneity of these products (Fig. 4B).
Evidence for the Presence of Araf: Composition of the Lipidlinked Polymers-The products recovered through extraction with CHCl 3 /CH 3  To provide further evidence for the presence of Ara in the lipid-linked polymer, the enzymatically active membranes and

FIG. 3. Gel filtration chromatography of the native and mild acid-hydrolyzed [ 14 C]Gal-labeled lipid-linked polymer.
A, the reaction mixture contained UDP-[ 14 C]Galp, dTDP-Rha, UDP-GlcNAc, 2 mg of membrane protein, 1.5 mg of cell envelope protein, 0.13 mg of mutase in buffer A in a total volume of 320 l. After a 1-h incubation at 37°C, the reaction mixture was subjected to the series of extractions. Some (300,000 cpm) of the E-soak extract was dried under a stream of N 2 and immediately dissolved in 600 l of 100 mM sodium acetate (pH 7.0) and loaded on a BioGel P-100 column (1 ϫ 118 cm) and eluted with 100 mM sodium acetate. Fractions of 1 ml were collected and counted. B, about 20,000 cpm each of the CHCl 3 /CH 3 OH/H 2 O and E-soak extracts were dried, suspended in 50 l of 1-propanol, sonicated in a bath sonicator, treated with 100 l of 0.02 N HCl at 60°C for 30 min, cooled, and neutralized with 10 l of 0.2 N NaOH. Sodium acetate (100 mM; 450 l) was added to each sample to achieve a volume of 600 l, which was applied to the BioGel P-100 column. cell envelope fraction were incubated with 1[ 14 C]-D-Araf-C 50 . About 2% of the input radioactivity was incorporated into combined CHCl 3 /CH 3 OH/H 2 O (10:10:3) and E-soak-soluble polymer and insoluble residue. Incorporation into E-soak-soluble lipid-linked polymer was twice that into CHCl 3 /CH 3 OH/H 2 Osoluble polymer, suggesting that the two families of lipid-linked polymers differed in the degree of arabinosylation. The nature of the E-soak-soluble [ 14 C]Ara-labeled polymer was examined after mild acid hydrolysis by gel filtration on BioGel P-100 and was shown to be similar in size to [ 14 C]Gal-labeled E-soak-soluble products.
The composition of the lipid-linked polymer was further examined by incorporation of 14 C from the individual sugar nucleotides, UDP-[ 14 C]GlcNAc, dTDP-[ 14 C]Rha, and the Ara precursor, 5-P-[ 14 C]ribosyl-PP, into the CHCl 3 /CH 3 OH/H 2 O (10: 10:3) and E-soak-soluble products. These and the [ 14 C]Gallinked products were compared by Tricine SDS-PAGE autoradiography. All show the same mobility pattern (Fig. 6A). Inclusion of the arabinan precursor, 5-P-[ 14 C]ribosyl-PP (16), in the reaction mixture, resulted in the appearance of even more glycosylated (presumably arabinosylated) products in Esoak-soluble material. Strong acid hydrolysis of the differently labeled lipid-linked polymer followed by TLC analysis of the released monosaccharides confirmed that the 14 C label was retained in the appropriate form, i.e. GlcNAc, Rha, or Gal, and that about 80% of the 14 C label derived from P-[ 14 C]ribosyl-PP in the lipid-linked polymer was converted to arabinose (Fig.  6B).
GL-1/GL-2 Are Precursors of the Lipid-linked Polymer-The products from a reaction mixture containing 60,000 cpm of the  Fig. 3. A, 500 cpm of native (lane 1) and acid-hydrolyzed (lane 2) [ 14 C]Gal-labeled E-soak-soluble products were applied to a silica gel TLC plate and developed in 60% methanol in water, containing 0.025% ammonium hydroxide. The plate was exposed to Kodak X-Omat AR film at Ϫ70°C for 7 days. B, 5000 cpm of [ 14 C]Gallabeled E-soak-soluble material (lane 1) and 5000 cpm of the CHCl 3 / CH 3 OH/H 2 O (10:10:3)-soluble material (lane 2) were dried, immediately dissolved in 10 l of SDS-sample buffer, boiled for 3 min, and loaded to a 15% SDS-polyacrylamide gel along with radiolabeled protein molecular weight markers provided as a gauge of relative mobility (lane 3). Blotting to nitrocellulose membrane was performed at 50 V for 1 h. The membrane was exposed to Kodak X-Omat AR film at Ϫ70°C for 2 days.

FIG. 5. Linkage analysis of per-O-methylated, per-O-acetylated [ 14 C]Gal-labeled E-soak-soluble lipid-linked polymer.
About 500,000 cpm of [ 14 C]Gal-labeled E-soak material were dried in vacuo and subjected to NaOH methylation as described (19). Per-O-methylated oligosaccharide alditol acetates were prepared, and ϳ150,000 cpm were analyzed by GC on a fused silica Durabond-1 column, coupled to the Lablogic GC-RAM radioactive counter. The individual peaks were identified by comparison of retention times of cold standard of permethylated, peracetylated acid solubilized arabinogalactan from M. bovis (5). The CHCl 3 /CH 3 OH/H 2 O-soluble products showed almost identical profiles. C]Gal. A, SDS-PAGE, followed by blotting to nitrocellulose was performed on the products of the above reactions. Aliquots representing 10% of the CHCl 3 /CH 3 OH/H 2 O and 5% of the E-soak extracts were dried, immediately dissolved in 10 l of SDS-sample buffer, boiled for 2 min, and loaded to a 10 -20% Tricine SDS-polyacrylamide gel along with radiolabeled protein molecular weight markers. The blot was exposed to Kodak BioMax MR film Ϫ70°C for 14 days.  8; 8000 cpm). In the image, lanes 4 and 8 from this autoradiogram were replaced with the same lanes from an autoradiogram that was exposed to film for 1 day. B, complete acid hydrolysis was conducted on the CHCl 3 /CH 3 OH/ H 2 O (10:10:3) and the E-soak extracts (ϳ1500 cpm) in 2 M CF 3 COOH for 1 h at 120°C. Hydrolysates (500 cpm) were chromatographed on silica gel plates in pyridine/ethyl acetate/glacial acetic acid/water (5:5:1:3) and developed twice. TLC plates were exposed to Kodak BioMax MR film at Ϫ70°C for 8 days. C, cold sugar standards were visualized by charring with 10% cupric sulfate in 8% phosphoric acid. The plate shown is from E-soak-soluble hydrolyzed polymer; however, the results from hydrolysis of the CHCl 3 /CH 3 OH/H 2 O (10:10:3)-soluble polymer were identical. purified GL-1/GL-2 mixture as the radioactive precursors were extracted with CHCl 3 /CH 3 OH/H 2 O and E-soak. Over 12% of the input radioactivity was incorporated into the final macromolecules, most of which was in the E-soak-extractable material. That these were the lipid-linked polymer was confirmed by DEAE-cellulose chromatography and mild acid hydrolysis followed by gel filtration. In another innovative approach to address the relationship between the simpler glycolipid and the lipid-linker polymers, a time course experiment was conducted using isolated membranes that had been prelabeled with UDP-[ 14 C]GlcNAc and then chased with cold UDP-Galp in the presence of the UDP-Galp mutase. A comparison of total radioactivity in the extracted fractions showed a decrease in the amount of CHCl 3 /CH 3 OH (2:1)-soluble precursors (i.e. [ 14 C] GlcNAc-labeled GL-1/GL-2) accompanied by an increase in radioactivity in the more polar, more glycosylated CHCl 3 / CH 3 OH/H 2 O and E-soak-soluble products (Fig. 7). Although counts lost did not equate fully with counts gained, which may be attributed to partial decomposition of the substrates, it is clear that conversion of the radiolabeled GL-1/GL-2 precursor to lipid-linked polymer occurred, probably due to endogenous glycosyl transferases present in the membrane and cell envelope fractions.
Cloning and Enzymatic Activity of the Galactosyltransferase Gene M. tuberculosis Rv3808c-Analysis of the genomic data base of M. tuberculosis H37Rv (21) revealed the gene Rv3808c downstream from the UDP-Galp mutase glf gene (Rv3809c). The first four nucleotides of Rv3808c and the last four of Rv3809c overlap. In light of this four-nucleotide overlap between RV3808c and glf, it seemed likely that Rv3808c comprises an operon with the glf (Rv3809c) gene, although both genes have possible ribosome binding sites. (This operon might extend up to Rv3805c, because Rv3807c, Rv3806c, and Rv3805c apparently do not have their own promoters.) Furthermore, hydrophobic cluster analysis of Rv3808c demonstrated a conserved ␤-glycosyltransferase domain (26). Thus, it seemed possible that Rv3808c was a galactosyltransferase transcription-ally coupled to glf. Rv3808c was cloned into the pET29b and pJJV7 expression vectors and transformed into E. coli and M. smegmatis cells. Overexpression of the gene in E. coli led to production of a new protein with an apparent molecular mass of 68 kDa, as determined by SDS-PAGE (data not shown) and as predicted. However, the observed level of expression was lower than normally obtained with this expression system, and all of the recombinant protein was found in the insoluble fraction of cell homogenates; no enzymatic activity was detected in the soluble portion of the homogenates. The mycobacterial expression plasmid pJJV7 differs in that expression is driven from a native, mycobacterial groEL promoter. M. smegmatis transformed with pJJV7-3808c plasmid or empty plasmid was examined for galactosyl transferase activity in the basic cellfree assay. A time course experiment showed almost no differences in the incorporation of [ 14 C]Gal from UDP-[ 14 C]Gal into the CHCl 3 /CH 3 OH (2:1) extracts; however, there was a substantial increase in radioactivity incorporated into CHCl 3 / CH 3 OH/H 2 O (10:10:3) and E-soak (Fig. 8). The former lipidlinked polymer was subjected to mild acid hydrolysis followed by gel filtration chromatography. The radioactive profile revealed that the products from the control strain and the strain with the cloned gene had the same size, and thus the increase in the incorporation of the [ 14 C]Gal into the products is not due to further extension of the galactan chain but rather due to greater production of the same material. Methylation analysis of the lipid-linked polymer confirmed that true galactofuran was synthesized in the reaction. Thus, Rv3808c encodes a galactosyl transferase responsible for the synthesis of bulk 5and 6-linked galactofuran. DISCUSSION To date, the only polyprenyl-P implicated in aspects of mycobacterial cell wall biosynthesis are decaprenyl-P and heptaprenyl-P (27,28). The addition of a cell wall-membrane enzyme preparation and UDP-[ 14 C]Galp to reaction mixtures capable of synthesizing polyprenyl-P-P-GlcNAc (GL-1) and polyprenyl-P-P-GlcNAc-Rha (GL-2) resulted in the synthesis of Galf-labeled more polar glycolipids, GL-3 and GL-4, indicating stepwise growth of the initial segments of the galactan chain on the polyprenyl-P-P-GlcNAc-Rha unit, 1 Gal unit at a time (7). Present evidence shows that thoroughly washed membrane preparations are not able to synthesize GL-3 and GL-4, which, however, could be achieved by the addition of UDP-Galp mutase encoded by the glf gene (11), demonstrating a requirement for UDP-Galf as donor. Moreover, analysis of the CHCl 3 /CH 3 OH (2:1)-soluble lipids in polar solvent demonstrated a hierarchial array of galactolipids, with all of the evidence for polyprenyl-P linkage, again pointing to sequential addition of single Galf units. Calling on the approaches that led to the solubilization of the oligosaccharide-P-P-dolichol intermediates of glycoprotein synthesis (13) and to the extraction of phosphosphingolipids from yeast (14) and the lipophosphoglycan of Leishmania donovani (29), we successfully solubilized the newly synthesized galactofuran. Surprisingly, two distinct populations exist, differentially extracted by the two solvents. As in the case of the dolichyl-bound oligosaccharides, the identification of a polyprenol-P linkage was based on mild acid lability, mild alkali stability, solubility in extremely polar organic solvents, and exclusion from Bio-Gel P-100, all suggesting a highly polymerized lipid-linked version of GL 1-4. De novo synthesis of the lipidlinked polymer is also sensitive to tunicamycin, and evidence is presented for the incorporation of GL-1/GL-2 into the lipidlinked polymer. Glycosyl linkage analysis of the polymer produced t-Galf, 5-linked Galf, 6-linked Galf, and 5,6-linked Galf, indicating that there is substitution of one or more of the linear Galf residues, presumably with arabinan. Moreover, [ 14 C]Araf donated by synthetic C 50 -P-[ 14 C]Araf, or formed from 5-phospho-[ 14 C]ribosyl-pyrophosphate was incorporated into this same polymer as characterized by solubility in polar lipid solvents, SDS-PAGE mobility, mild acid lability, and hence lipid linkage. The combined evidence points to the pathway shown in Fig. 9 for the synthesis of the AG component of the mycolylarabinogalactan complex of mycobacterial cell walls. Clearly, this cell-free system does not allow for significant transfer of the newly synthesized AG from polyprenyl-P-P to peptidoglycan, in that relatively little of the [ 14 C]Gal from UDP-[ 14 C]Gal ends up in the insoluble mycolylarabinogalactan.
The search for galactosyl transferases responsible for this galactan AG elongation through comparisons with various families of galactopyranosyl transferases (30) proved to be uninformative. However, analysis of M. tuberculosis Rv3808c, which is linked directly to the glf gene (Rv3809c), showed strong indications of a glycosyl transferase. Alignment of the hydrophobic cluster analysis plot (31) of the predicted amino acid sequence of M. tuberculosis Rv3808c with plots of other known ␤-glycosyltransferases (26) revealed a common domain structure of repeating ␣-helix and ␤-strand motifs between amino acids 161 and 262, corresponding to domain A of glycosyltransferases (31). Within this domain, two aspartic acid residues at 199 and 256 had the hallmarks of highly conserved residues within the C-terminal loops of the ␤-2 and ␤-4 strands, the characteristic signature of all ␤-glycosyltransferases (26). There was no evidence of domain B in Rv3808c (and no conserved QXXRW amino acid motif) characteristic of ␤-glycosyltransferases that add sugars processively to the reducing end of a polysaccharide chain. ␤-Glycosyltransferases that add sugar residues to the nonreducing end of the polysaccharide chain have only the one domain, A (26). Thus, the synthesis of mycobacterial galactan may share similarities with biosynthesis of the homopolymer O-antigenic D-galactans of E. coli O8 and O9 and Klebsiella pneumoniae O1. In both cases, synthesis is initiated by the transfer of GlcNAc-P to polyprenyl-P (32), and, during formation of the E. coli O9 antigen, mannosyl residues are rapidly added to the nonreducing terminus of the acceptor one residue at a time, in a processive mechanism (33). D-Galactan I from K. pneumoniae is assembled on the cytoplasmic face of the plasma membrane, and polymerization is thought to occur by sequential sugar transfer on the lipid intermediate. An ATP-binding cassette (ABC) transporter then translocates polymerized Dgalactan I across the plasma membrane prior to ligation to lipid A core (33).
The elucidation of the basic elements of synthesis of the cell wall core of mycobacteria should substantially enhance current tuberculosis drug discovery efforts, in that aspects of cell wall synthesis are the targets of many of the current front-line anti-tuberculosis drugs (2), and the pathways and their end products are distinctly xenogeneic. FIG. 9. Pathway for the early steps in the synthesis of the arabinogalactan heteropolysaccharide of mycobacterial cell wall core. The genes encoding the galactosyltransferases have not yet been defined, except for the putative galactosyl transferase, M. tuberculosis Rv3808c gene product, identified in this study. The arabinosyltransferases may be encoded by the ethambutol resistance genes, embA to -C (34). The values for m, n, x, and y are not known.