JOURNAL OF BACTERIOLOGY, Dec. 1995, p. 7125–7130 0021-9193/95/$04.0010 Copyright q 1995, American Society for Microbiology
Vol. 177, No. 24
Biosynthetic Origin of Mycobacterial Cell Wall Arabinosyl Residues MICHAEL SCHERMAN,1 ANTHONY WESTON,2 KEN DUNCAN,2 ANDY WHITTINGTON,2 RICHARD UPTON,2 LINGYI DENG,1† ROBERT COMBER,3‡ JOYCE D. FRIEDRICH,3 1 AND MICHAEL MCNEIL * Department of Microbiology, Colorado State University, Fort Collins, Colorado 805231; Glaxo Research and Development Ltd., Stevenage, Hertfordshire, United Kingdom2; and Southern Research Institute, Birmingham, Alabama 352553 Received 21 June 1995/Accepted 2 October 1995
Designing new drugs that inhibit the biosynthesis of the D-arabinan moiety of the mycobacterial cell wall arabinogalactan is one important basic approach for treatment of mycobacterial diseases. However, the biosynthetic origin of the D-arabinosyl monosaccharide residues themselves is not known. To obtain information on this issue, mycobacteria growing in culture were fed glucose labeled with 14C or 3H in specific positions. The resulting radiolabeled cell walls were isolated and hydrolyzed, the arabinose and galactose were separated by high-pressure liquid chromatography, and the radioactivity in each sugar was determined. [U-14C]glucose, [6-3H]glucose, [6-14C]glucose, and [1-14C]glucose were all converted to cell wall arabinosyl residues with equal retention of radioactivity. The positions of the labeled atoms in the arabinose made from [1-14C]glucose and [6-3H]glucose were shown to be C-1 and H-5, respectively. These results demonstrated that the arabinose carbon skeleton is formed via the nonoxidative pentose shunt and not via hexose decarboxylation or via triose condensations. Since the pentose shunt product, ribulose-5-phosphate, is converted to arabinose-5-phosphate as the first step in 3-keto-D-manno-octulosonic acid biosynthesis by gram-negative bacteria, such a conversion was then searched for in mycobacteria. However, cell-free enzymatic analysis using both phosphorous nuclear magnetic resonance spectrometry and colorimetric methods failed to detect the conversion. Thus, the conversion of the pentose shunt intermediates to the D-arabino stereochemistry is not via the expected isomerase but rather must occur via novel metabolic transformations. (7a, 21). It remains unknown if b-D-Araf-monophosphodecaprenol is the sole donor of arabinosyl residues or if there is an aqueous soluble arabinosyl nucleotide donor as well. In this regard, a preliminary report of a partially purified uridine nucleotide of arabinose in mycobacteria (13) has appeared. In this study, we turn our attention to the early metabolic events of arabinan biosynthesis and address the basic issue of how the carbon skeleton of the arabinosyl residues is formed. This research has been frustrated by our inability to prepare a cell-free system which reproducibly converts general precursors of arabinose, such as radiolabeled glucose, to arabinosyl residues. However, in contrast to broken cells, whole bacteria do convert D-[U-14C]glucose into cell wall arabinan. Advantage was taken of this fact by labeling with glucose radioactive in specific carbons and hydrogens and analyzing the resulting arabinosyl residues for radioactivity. The results of these labeling experiments were then used to deduce how the arabinose skeleton is formed. In addition, mycobacterial enzyme extracts were assayed for the presence of relevant isomerase and mutase activities.
The structure (Fig. 1) of the mycobacterial cell wall arabinan moiety of the arabinogalactan (henceforth referred to simply as arabinan) reveals that it is a very complex branched polysaccharide of arabinofuranosyl (Araf) resides (1). It provides a key structural connection between the peptidoglycan and the mycolic acids (9). Thus, the mycolic acids are attached in a specific fashion to a nonreducing end pentaarabinofuranoside [b-D-Araf-(132)-a-D-Araf-(13]2-3&5)-a-D-Araf (11). The reducing end of the arabinan is attached to a galactofuran (1) which is finally attached to the peptidoglycan via the actinomycete-specific linker disaccharide [34-Rha-(133)-GlcNAc(13phosphate)] (10). This structural information suggests that arabinan is essential for cell wall integrity, and the antimycobacterial activity of ethambutol, an inhibitor of arabinan formation (3, 14), confirms this expectation. In parallel with what is known about the biosynthesis of other polysaccharides (12), it is expected that the arabinan is synthesized via arabinosyltransferases. These enzymes should utilize a 1-phosphorylated arabinose (either an arabinosyl sugar nucleotide or an arabinosyl polyprenyl phosphate) as the donor and the incomplete growing arabinan as the acceptor. Indeed, b-D-Araf-monophosphodecaprenol has been isolated and characterized from a variety of mycobacteria (18, 19), and it has been demonstrated to function as an arabinosyl donor
MATERIALS AND METHODS Bacterial strain, media, and chemicals. Mycobacterium smegmatis NCIB 8548 was used for all labeling experiments and was grown as described below. M. smegmatis ATCC 607 was used to prepare active enzyme extracts. For these purposes, it was grown in a 40-liter fermentor in GA medium containing asparagine (5 g/liter), potassium dihydrogen phosphate (5 g/liter), glycerol (25 g/liter), MgSO4 z 7H2O (1.3 g/liter), and trace salts. The pH was adjusted to 6.8 prior to autoclaving. The culture was grown to late exponential phase, cells were harvested by centrifugation, and cell pellets were stored at 2708C. For control experiments with Escherichia coli, the strain E. coli B was used. It was grown in L broth containing 1% glucose at 378C with shaking to an optical density at 600 nm of approximately 1.0. Cells were harvested by centrifugation, and the pellets were stored at 2708C. D-[U-14C]glucose (260 Ci/mol), D-[1-14C]glucose (55 Ci/
* Corresponding author. Mailing address: Department of Microbiology, Colorado State University, Fort Collins, CO 80523. Phone: (970) 491-1784. Fax: (970) 491-1815. Electronic mail address: mmcne il@
[email protected]. † Present address: Channing Laboratory, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA 02115. ‡ Present address: Carson Products Company, Savannah, GA 31403. 7125
7126
SCHERMAN ET AL.
FIG. 1. Structure of the mycobacterial cell wall arabinan. Its relationship to the other major components of the mycobacterial cell wall is also illustrated.
mol), D-[6-14C]glucose (55 Ci/mol), D-[6-3H]glucose (30,000 Ci/mol), and D-[114 C]ribose (55 Ci/mol) were purchased from American Radiolabelled Chemicals, Inc. (St. Louis, Mo.). D-Arabinose-5-phosphate (Ara-5-P), a-D-ribofuranose-1phosphate (Ribf-1-P), D-ribulose-5-phosphate (Ru-5-P), and D-ribose-5-phosphate (Rib-5-P) were purchased from Sigma (St. Louis, Mo.). a-D-Arabinofuranose-1-phosphate and a-D-arabinopyranose-1-phosphate were synthesized by Malcolm Berry at Glaxo Research and Development by the methods of Wright and Khorana (20). b-D-Arabinofuranose-1-phosphate was synthesized as described previously (19). Treatment of bacteria with radiolabeled glucose and ribose. M. smegmatis (NCIB 8548) was inoculated into 200 ml of nutrient broth (Sigma) and grown to mid-log phase at 378C with agitation. Then 2.0 ml of this culture was aliquoted into test tubes for labeling. One to 10 mCi of a radiolabeled glucose was added along with 18 nmol of nonradioactive glucose. The tubes were then allowed to incubate for 10 min, after which 10 mmol of nonradioactive glucose was added. The tubes were then incubated for an additional 2 h, and the cells were harvested by centrifugation. The whole cell pellet in each tube was washed several times with water and resuspended in phosphate-buffered saline (PBS; pH 6.8) containing 0.5% (vol/vol) Triton X-100 detergent (PBS-Triton). This mixture was then sonicated for a total of 4 min until nearly complete cell breakage had occurred, as verified by acid-fast staining and microscopy. The solution was then centrifuged at 14,000 3 g for 10 min. The pellet was washed once more in the PBS-Triton and twice more in water to yield the radiolabeled cell wall. Radiolabeling with ribose was done in an analogous fashion, using D-[1-14C]ribose and nonradioactive ribose in place of the glucose compounds. Hydrolysis of cell wall pellets, separation of monosaccharides, and determination of radioactivity in each. The cell wall pellets were dried under a gentle stream of air in a heated sand bath, taken up in 0.3 ml of 2 M trifluoroacetic acid, and heated at 1158C for 60 min. Then the solution was blown to dryness and subsequently subjected to two methanol blowdowns, to help remove any residual trifluoroacetic acid. The pellets were taken up in about 50 ml of water, and 20 nmol each of D-glucose, D-galactose, D-mannose, D-ribose, L-rhamnose, and D-arabinose was added. This mixture was then injected onto a Dionex highpressure liquid chromatography (HPLC) system (Dionex, Sunnyvale, Calif.) with a CarboPack PA1 anion-exchange column measuring 4 by 250 mm and a pulsed amperometric detector. The column was eluted at 1 ml/min with an isocratic elution of 100 mM NaOH. A postcolumn addition of 0.3 ml of 0.3 M NaOH per min was done to aid pulsed amperometric detection. Thirty fractions were collected at 1-min intervals. An aliquot of each fraction was counted on a liquid scintillation system, and the resulting radioactivity was plotted against time. Location of radioisotope in monosaccharides. Advantage was taken of the fact that NaIO4 treatment of a monosaccharide cleaves between all of the backbone carbon atoms. The hydroxy methyl group (at position 6 in Gal and at position 5 in Ara) is converted to formaldehyde, while the secondary CHOH groups and the CHO group at position 1 are converted to formic acid. In contrast, if the monosaccharide is reduced with NaBH4 before NaIO4 treatment, thus converting the CHO group at position 1 to a hydroxy methyl group, then both ends of the molecule (positions 1 and 5 in Ara) are converted to formaldehyde and the internal CHOH groups are converted to formic acid. The formaldehyde and formic acid are readily separated by ion-exchange chromatography. Hence, by
J. BACTERIOL. performing NaIO4 oxidation on both untreated and NaBH4-reduced Ara and Gal, the percentage of isotope at position 5/6, position 1, and the internal positions can be determined. Therefore, radiolabeled cell wall arabinose and galactose purified by HPLC were passed through Dowex-50 resin in the hydrogen form to desalt. Each sample was then split in two. One part was reduced by treatment with 50 ml of sodium borohydride at 10 mg/ml in a 1:1 mixture of ethanol and 1 M NH4OH for 1 h. Borate was removed by codistillation with CH3OH after neutralization with acetic acid as described previously (22). Separately, the reduced and nonreduced samples were treated with 150 ml of 0.5 M NaIO4 in 0.5 M sodium phosphate buffer (pH 5.8) for 1 h at room temperature in the dark. The reaction mixtures were passed through a 200-ml column of Dowex 1 (8% cross-linkage, 200 by 400 mesh; Sigma) in the hydroxide form and then washed with water. The combined effluents were counted and regarded as the neutral fractions. The columns were then washed with 1 M HCl, and the effluent was counted and regarded as the acidic fractions. The amount of isotope at the primary hydroxyl methyl group (position 5 in Ara and position 6 in Gal) was calculated by using only the non-NaBH4-reduced sample data according to the following formula: percentage of isotope at hydroxy methyl 5 100 3 [cpmneutral/(cpmneutral 1 cpmacidic)]. The amount of isotope at positions 1 and 5/6 together was calculated by using only the data from the NaBH4-reduced sample in the same way. With this information, the percentage of label at position 1 and at the internal positions was readily deduced. The validity of the method was verified by using [1-14C]Glc, [3,4-14C]Glc, and [6-3H]Glc. Preparation of cell extracts. M. smegmatis cells were resuspended in 50 mM Tris HCl buffer (pH 8.0) plus 10 mM MgCl2 to a concentration of 50% (wt/vol), sonicated at 48C for 30 s in an Ultrasonic sonicator, and then cooled for 90 s. This procedure was repeated six times. Unbroken cells were removed by centrifugation in a Beckman JA-20 rotor at 17,600 3 g for 15 min. The supernatant was recentrifuged at 25,400 3 g for 15 min. The supernatant was used as the crude enzyme extract, and the protein concentration was routinely between 20 and 25 mg/ml. In the case of E. coli, cells (16.5 g) were resuspended in 50 mM phosphate buffer (pH 7.2) containing 0.5 mM dithiothreitol, 1 mM EDTA, and 1 mM MgCl2 at 48C to a final volume of 50 ml. The cells were broken by three passages in a French press, and the cell debris was removed by centrifugation at 27,000 3 g. The supernatant was centrifuged at 100,000 3 g for 1 h, and the supernatant was used as the crude soluble enzyme extract. Assay for arabinosephosphate isomerase, ribosephosphate isomerase, phosphoribomutase, and a-D-arabinose-1-phosphate epimerase. All assays were based in the carbazole color reaction specific to ketose phosphates described by Volk (16). Thus, for arabinosephosphate isomerase, the substrate used was Ara-5-P; for ribosephosphate isomerase, the substrate was Rib-5-P. The expected product in both cases was Ru-5-P, which gives a positive carbazole reaction. The 5-phosphoribose mutase was assayed by recognition that the product of the mutase acting on Ribf-1-P would be Rib-5-P, which was by then known to be converted to Ru-5-P (see Table 2). In a similar fashion, the presence or absence of a-Araf-1-P epimerase was assayed, knowing that Ribf-1-P was converted by the combined action of the mutase and isomerase to Ru-5-P. The enzymes to perform the assay from M. smegmatis crude cell extracts were freed from small molecules on a Pharmacia PD10 column in 50 mM Tris HCl buffer (pH 8.0) containing 10 mM MgCl2. Protein fractions were pooled to give a final protein concentration of 10 mg/ml. Sixty microliters of 20 mM sugar phosphate (Ara-5-P, Rib-5-P, Ribf-1-P, or a-Araf-1-P) was mixed with various amounts (see Table 2) of M. smegmatis extract, and the assay mixture was made up to 300 ml with buffer. The assay mixtures were incubated at 378C for 1 h, after which time 0.5 ml of 0.1 M borate buffer (pH 8.0) was added. This was followed immediately by the addition of 3.0 ml of concentrated H2SO4, 0.1 ml of 1.5% cysteine, and 0.1 ml of 0.12% carbazole in 95% ethanol. The amount of keto sugar phosphate (presumed to be Ru-5-P) was measured at 590 nm. For the E. coli controls, the soluble protein extract was subjected to a Pharmacia PD10 column and the protein was eluted in 0.1 M glycylglycine buffer (pH 8.0) to remove free sugar phosphates. Pooled fractions had a final protein concentration of 1.9 mg/ml. Incubation with Ara-5-P was as described above except that the incubation time was only 30 min. Assay for phosphoarabinomutase by HPLC and 31P NMR spectrometry. A crude extract of M. smegmatis (23 mg/ml) was dialyzed against 50 mM Tris HCl (pH 8.0) containing 10 mM MgCl2 overnight at 48C. Four hundred microliters of the extract was mixed with 0.1 ml of Ara-5-P (20 mg/ml) and incubated at 378C for 6 h. An equal volume of absolute ethanol was added, and after 10 min at room temperature, the precipitated protein was removed by centrifugation. The sample was then analyzed by HPLC and by 31P nuclear magnetic resonance (NMR) spectrometry. For HPLC, the sample was diluted 1:80 with water and a sample (20 to 50 ml) was injected on a Dionex PA100 column equilibrated in 100 mM sodium acetate in 100 mM NaOH. Detection was performed with the Dionex amperometric detector. The column flow was 1 ml/min. A gradient of sodium acetate in the 100 mM NaOH was as follows: 100 to 200 mM from t 5 0 to t 5 20; 200 to 500 mM from t 5 20 to t 5 30; and a 500 mM wash from t 5 30 to t 5 40. For 31P NMR spectrometry, 0.5 ml of sample along with 0.25 ml of 2H2O (Aldrich) was run on a Varian VXR400 spectrometer operating at 161.9 MHz (31P) at 308C.
VOL. 177, 1995
ARABINAN BIOSYNTHESIS IN MYCOBACTERIA
7127
RESULTS Conversion of [U-14C]glucose into mycobacterial cell wall arabinan. Conditions for successful conversion of [U-14C]glucose into cell wall arabinan by growing cultures of M. smegmatis were obtained. Thus, growing cultures of M. smegmatis were treated with [U-14C]glucose for various periods of time, the label was chased to the cell wall by treatment with a high concentration of unlabeled glucose, the cells were recovered, the cell wall was isolated and hydrolyzed, and the resulting arabinose and galactose were separated by HPLC and counted. The complex medium, nutrient broth, resulted in a more reproducible and facile conversion of radiolabeled glucose to cell wall than the defined medium, glycerol-alanine-salts (1). Significant radioactivity in the cell wall could be obtained by using the conditions detailed in Materials and Methods in as little as 5 min; an incubation of 10 min was chosen as a compromise between a short labeling time to avoid nonspecific label scrambling and adequacy of radioactivity converted into cell wall arabinan. It should be noted that under these conditions, a reproducible ratio of label into cell wall arabinan and galactan was readily obtained (Ara/Gal ratio of ;1.5:1). However, the absolute yield of radioactivity into the cell wall varied between experiments performed on different days, most likely because of the age and condition of the cultures and perhaps also the absolute recovery of cell wall. Radioisotopes present at C-6 and at C-1 of glucose were incorporated in cell wall arabinan with equal efficiencies. The HPLC analyses of the arabinose and galactose released from mycobacterial cell walls labeled with [U-14C]glucose, [1-14C] glucose, and [6-14C]glucose are presented in Fig. 2. Somewhat unexpectedly, the 14C atoms were converted to cell wall arabinan in comparison to cell wall galactan with equal efficiencies regardless of their positions in the starting [14C]glucose. To confirm this finding, a double-labeling experiment using [1-14C]glucose and [6-3H]glucose was conducted. Equal counts per minute (but different disintegrations per minute because of the difference in counting efficiencies) of [1-14C]glucose and [6-3H]glucose were added at the same time to the M. smegmatis culture. The cell walls were isolated and hydrolyzed, the resulting sugars were separated by HPLC, and the radioactivity due to each isotope was determined (Fig. 3). This finding confirmed that label at position 1 and label at position 6 of glucose are converted to cell wall arabinan with equal efficiencies and ruled out formation of the arabinose carbon skeleton via loss of either the C-6 or C-1 of glucose or any other hexose. Positions of the labeled atoms in the mycobacterial M. smegmatis-synthesized arabinan were found to be H-5 when [6-3H] glucose was used as a label and mostly C-1 when [1-14C]glucose was used as a label. The results presented above are most straightforwardly reconciled with the arabinosyl carbon skeleton being formed via the nonoxidative pathway of the pentose shunt. If this pathway is utilized, five hexoses would be converted to six pentoses and all carbon atoms regardless of position (or hydrogen atoms originating at C-6) would be converted to pentose and ultimately to arabinan. However, the results are also consistent with breakdown of the glucose skeleton via glycolysis to triose phosphates and subsequent buildup back to pentose and ultimately arabinan. The two possibilities can be distinguished by analysis of the isotope positions in the radiolabeled arabinose molecules. Examination of the pathway for the formation of six pentoses from five hexoses via the pentose shunt reveals that 100% of the label at H-6 should be found at H-5 of the arabinose; 80% of the label originating at C-1 of glucose should be found at C-1 of arabinose, with
FIG. 2. Dionex HPLC chromatography of the radioactive sugars released from mycobacterial cell walls isolated after incubation of cultures of M. smegmatis with radiolabeled glucose. Labeling, isolation of cell walls, hydrolysis, and HPLC were performed as described in Materials and Methods. (A) Labeling with [U-14C]glucose; (B) labeling with [1-14C]glucose; (C) labeling with [6-14C] glucose.
20% at C-5. In contrast, if pentose is formed from triose phosphate, the position of the nuclides will be dependent on which triose, dihydroxyacetone phosphate or glyceraldehyde-3-phosphate, is built back into pentose but independent of whether the
FIG. 3. Dionex HPLC chromatography of the radioactive sugars released from mycobacterial cell wall after incubation of cultures of M. smegmatis with equal counts per minute of [1-14C]glucose and [6-3H]glucose. Labeling, isolation of cell walls, hydrolysis, and HPLC were performed as described in Materials and Methods. Standard scintillation counting techniques were used to separately quantitate radioactivity from 14C and 3H.
7128
SCHERMAN ET AL.
J. BACTERIOL.
TABLE 1. Locations of radionuclide in arabinose and galactose purified from [1-14C]glucose- and [6-3H]glucose-labeled mycobacterial cell wallsa
Labeled with [114 C]glucose
Arabinose Galactose
Amt (ml) of: Substrate (4 mM)
% of nuclide at indicated location Sugar
TABLE 2. Cell-free assay for the conversion of various pentose phosphates to keto sugar phosphates
Labeled with [6-3H]glucose
Ara-5-P
C-1
Internal carbons
C-5/6
H-1
Internal hydrogens
H-5/6
61 63
14 11
25 26
2 3
6 8
92 88
Rib-5-P a-Ribf-1-P
a
M. smegmatis was incubated with the indicated radiolabel, cells were harvested, and cell walls were prepared and hydrolyzed. The resulting arabinose and galactose were purified by HPLC, and the radionuclide was located.
a-Araf-1-P
M. smegmatis extracta
E. coli extracta
0 200 0 0 50 0 100 0 200
0 0 100 0 0 0 0 0 0
A590 in carbazole
0.004 0.004 0.241 0.004 0.724 0.003 0.496 0.003 0.003
a Enzyme extracts were prepared and substrates were incubated as described in Materials and Methods.
nuclide originates at position 6 or 1 of the starting radioactive glucose. This is true because dihydroxyacetone phosphate and glyceraldehyde-3-phosphate are readily interconverted, thus making label that originally originated at one or the other end of the glucose always at the 3 position in both trioses. Accordingly, both nuclides from the double-labeling experiment using [1-14C]glucose and [6-3H]glucose will be found at position 1 of the arabinose if the triose utilized is dihydroxyacetone phosphate or at position 5 if the triose utilized is glyceraldehyde-3-phosphate. Hence, the label in both the arabinose and, for a control, the galactose were determined by using the radiolabeled sugars formed by acid hydrolysis of cell walls doubly labeled with [1-14C]glucose and [6-3H]glucose. The results (Table 1) were clearly in accordance with the prediction of the arabinose skeleton being formed by the nonoxidative pentose shunt pathway, as the tritium was almost exclusively at H-5 and the 14C was mostly at C-1. Cell wall arabinan was preferentially labeled with respect to galactan when M. smegmatis cells were incubated with [1-14C] ribose. The conclusion that the arabinose carbon skeleton is formed via the nonoxidative pentose shunt and not via triose or hexose decarboxylation suggested that labeling with a pentose shunt intermediate should preferentially label cell wall arabinan rather than cell wall galactan. Thus, M. smegmatis cultures were incubated with [1-14C]D-ribose in the expectation that this sugar would be taken up and converted either to Rib-5-P (a pentose shunt intermediate) by direct phosphorylation or to Ru-5-P after isomerization (4) and phosphorylation. Indeed, M. smegmatis was able to convert 14C atoms from ribose into cell wall, and analysis of these cell walls (Fig. 4) showed that, as predicted, the label was preferentially incorporated into cell wall arabinan.
No arabinosephosphate isomerase is detectable in mycobacterial extracts. Given the fact that the arabinosyl carbon skeleton originates in the pentose shunt, it was appropriate to check the hypothesis put forth by several researchers (8, 14) that the D-arabino configuration results from the action of arabinosephosphate isomerase. To this end, enzyme extracts of M. smegmatis and E. coli were prepared and tested for arabinosephosphate isomerase and ribosephosphate isomerase activities. The results (Table 2) showed ribosephosphate isomerase activity but no arabinosephosphate isomerase activity in M. smegmatis. Arabinosephosphate isomerase was found in E. coli, which was used as a control (Table 2). Additionally, the enzyme phosphoribomutase could be detected in M. smegmatis as Ribf-1-P was converted to ketose, presumably after conversion by the mutase to Rib-5-P. Finally, these experiments showed that a-Araf-1-P was not epimerized to a-D-Ribf-1-P, since a-D-Ribf-1-P, if formed, would have yielded a positive reaction (Table 2). No phosphoarabinomutase is detectable in mycobacterial extracts. We also assayed for the presence of a phosphoarabinomutase which would convert Ara-5-P to a- or b-Araf-1-P, a possible substrate for arabinose sugar nucleotide formation (14). Thus, the conversion of Ara-5-P to a- or b-Araf-1-P was investigated by HPLC analysis. The results (Fig. 5) showed that incubation of Ara-5-P with mycobacterial enzymes did not result in production of either Araf-1-P. 31P NMR analysis (Fig. 6) confirmed this conclusion and further showed that Ara-5-P was metabolized by the mycobacterial enzymes into Pi, explaining the loss of starting material observed by HPLC (Fig. 5). An additional phosphorylated component also appeared in the 31P NMR analysis (Fig. 6). However, this compound does not appear in the HPLC analysis (Fig. 5) and thus does not originate from any pentose or pentulose phosphate. DISCUSSION
FIG. 4. Dionex HPLC chromatography of the radioactive sugars released from mycobacterial cell wall after incubation of cultures of M. smegmatis with [1-14C]ribose. This selective labeling of arabinose should be contrasted with that shown for labeling with [U-14C]glucose as shown in Fig. 2. Radiolabeling, isolation of cell walls, hydrolysis, and HPLC were performed as described in Materials and Methods.
Confidence that results obtained in assays using the fastgrowing M. smegmatis are applicable to the slowly growing pathogenic M. tuberculosis and M. avium as well as the nonculturable M. leprae is afforded in the fact that a detailed study of their cell walls demonstrated that the arabinans were indistinguishable (2). It is therefore highly likely that the arabinan biosynthetic pathways are also the same in all mycobacteria. The carbon skeletons of other pentoses such as D-xylose (12), L-arabinose (12), and D-apiose (6) are formed via C-6 decarboxylation of UDP-Glc. However, the results of the labeling experiments presented here rule out C-6 decarboxylation of a
VOL. 177, 1995
ARABINAN BIOSYNTHESIS IN MYCOBACTERIA
7129
pentose phosphates goes on to form the D-arabino stereochemistry. We have shown here that Ru-5-P is not isomerized to Ara-5-P as is found in organisms that require arabinose-5-phosphate for 2-keto-D-manno-octulosonic acid biosynthesis (8, 15). A negative result needs to be regarded with some care, but we point out that we also could not detect the presence of the next putative enzyme in a pathway utilizing arabinosephosphate isomerase, the mutase which would convert Ara-5-P to Araf-1-P. In addition, the ribosephosphate isomerase and phosphoribomutase were active in the M. smegmatis enzyme preparation, and we were able to show arabinosephosphate isomerase activity in another organism. Conversion of xylulose-5-phosphate to the D-arabino stereochemistry would require both an isomerization and epimerization, and thus it seems more likely that the D-arabino stereochemistry results from a C-2 epimerization of a ribosyl compound. The results presented in Table 1 show that mycobacteria can convert Rib-5-P to a-D-Ribf-1-P. However, we have also shown that a-D-Araf-1-P is not converted to a-D-Ribf-1-P, and thus the epimerization does not take place at the level of a-D-Ribf-1-P. Wolucka (18) has recently suggested that the epimerization could take place at the pentose lipid level, with b-D-Ribfmonophosphodecaprenol being epimerized to form b-D-Arafmonophosphodecaprenol. The data presented herein are consistent with such a hypothesis but do not prove that this occurs. Alternatively, the epimerization could occur at the sugar nucleotide level, in a fashion analogous to formation of UDP-Gal from UDP-Glc (12). This scenario could involve the epimerization of NDP-Ribf or the epimerization and ring contraction of NDP-Ribp. The nucleoside diphosphate (NDP)-Ribf could be formed from a-D-Ribf-1-P, which we have shown to be produced by a Rib-5-P mutase (Table 2); NDP-Ribp would require the formation of a ribopyranose-1-phosphate. Another possibility involving an arabinosyl nucleotide would be the
FIG. 5. Dionex HPLC chromatography of the sugar phosphates produced before (A) and after (B) incubation of Ara-5-P with an enzyme extract prepared from M. smegmatis. Neither a- nor b-Araf-1-P was formed, although the amount of Ara-5-P decreased. The retention times of the standards a-Arap-1-P (a), a-Ribf-1-P (b), a-Araf-1-P (c), Rib-5-P (d), and Ru-5-P (e) are indicated in panel B. b-Araf-1-P elutes slightly in front of a-Araf-1-P.
uronic acid sugar nucleotide as the route of formation of the D-arabinosyl carbon skeletons. The results also rule out condensations utilizing triose. In contrast, the data for both radiolabeled glucose and radiolabeled ribose are consistent with the arabinosyl carbon skeleton being formed via the pentose shunt. Surprisingly, at least under the conditions used, M. smegmatis utilizes only the nonoxidative pentose shunt and does not utilize the oxidative pathway converting glucose-6-phosphate to Ru-5-P. This is true because utilization of the oxidative pathway would have resulted in the loss of C-1 from glucose. Since the oxidative pathway of the pentose shunt is normally the major pathway for NADPH formation and since NADPH is utilized in mycolic acid formation (7), it would have been reasonable to find that the oxidative leg of the pentose shunt is highly utilized for the dual purpose of supplying the bacteria with the large amounts of both pentose and NADPH needed for cell wall biosynthesis. The three end pentoses formed by the conversion five hexose-6-phosphates to six pentose-5-phosphates are Rib-5-P, Ru-5-P, and xylulose-5-phosphate. Thus, one of these three
FIG. 6. 31P NMR spectrometry before (A) and after (B) incubation of Ara5-P with an enzyme extract prepared from M. smegmatis. The loss of Ara-5-P is evident, as is the appearance of Pi and the appearance of a new compound (x) which was not identified. However, no pentofuranose-1-phosphate at 2.4 ppm was formed. The chemical shifts of the phosphate groups of Ru-5-P (a), Rib-5-P (b), and a-Ribf-1-P (c) are indicated in trace B.
7130
SCHERMAN ET AL.
formation of arabinose-1-phosphate via some mechanism other than a mutase acting on Ara-5-P followed by reaction with an NTP to form the arabinosyl nucleotide. Key relevant information for distinguishing between these possibilities is determining the presence or absence of ribo and/or arabino sugar nucleotides in the two ring forms in mycobacteria. In the case of ribose, such a sugar nucleotide is not to be confused with ADP-ribose, a fundamentally different compound in that the ribose is attached to the ADP at its 5 rather than 1 position. The unexpected scenario of no arabino or ribo sugar nucleotide should also be considered. In this case, an RNA precursor such as 5-phosphoribose pyrophosphate would be the substrate for epimerization and/or conversion to the phosphodecaprenyl pentose (5). It should also be noted that mycobacteria can be forced to grow on D-arabinose (17). In this situation, the arabinose is reduced to arabinitol and reoxidized to xylulose, which is then presumably phosphorylated. This important observation probably has more to do with the utilization of an unusual carbon source by mycobacteria than the synthesis of arabinan. In conclusion, from the studies herein, we now know the arabinosyl carbon skeleton is formed via the pentose shunt and that the obvious pathway from the pentose shunt via arabinosephosphate isomerase is not utilized. Determination of the pathway used by mycobacteria to convert a pentose shunt intermediate to b-D-Araf-monophosphodecaprenol is the next logical step to acquire fundamental information needed to fully exploit the potential of targeting arabinan biosynthesis for the development of new drugs. ACKNOWLEDGMENTS This work was supported by funds provided by Public Health Service grant NIAID NIH (AI-33706). We gratefully acknowledge the fundamental contributions of Patrick Brennan to these metabolic studies. We also acknowledge and appreciate the discussions with and work of Bill McDowell, Brian Carter, Paul Smith, and Malcolm Berry. REFERENCES 1. Daffe, M., P. J. Brennan, and M. McNeil. 1990. Predominant structural features of the cell wall arabinogalactan of Mycobacterium tuberculosis as revealed through characterization of oligoglycosyl alditol fragments by gas chromatography/mass spectrometry and by 1H and 13C-NMR analyses. J. Biol. Chem. 265:6734–6743. 2. Daffe, M., M. McNeil, and P. J. Brennan. 1993. Major structural features of the cell wall arabinogalactans of Mycobacterium, Rhodococcus, and Nocardia spp. Carbohydr. Res. 249:383–398. 3. Deng, L., K. Mikusova, K. G. Robuck, M. Scherman, P. J. Brennan, and M.
J. BACTERIOL. McNeil. 1995. Recognition of multiple effects of ethambutol on the metabolism of the mycobacterial cell envelope. Antimicrob. Agents Chemother. 39:694–701. 4. Izumori, K., A. W. Rees, and A. D. Elbein. 1975. Purification, crystallization, and properties of D-ribose isomerase from Mycobacterium smegmatis. J. Biol. Chem. 250:8085–8087. 5. Kalbe-Bournonville, L. 1994. Unique D-arabinosyl and myo-inositol compounds synthesized by mycobacteria suggest unusual biosynthetic pathways. Ph.D. thesis. Universite´ Catholique de Louvain, Brussels, Belgium. 6. Kindel, P. K., and R. R. Watson. 1973. Synthesis, characterization and properties of uridine 59(a-D-apio-D-furanosyl pyrophosphate). Biochem. J. 133:227–241. 7. Lacave, C., A. Quemard, and G. Laneelle. 1990. Cell-free synthesis of mycolic acids in Mycobacterium aurum: radioactivity distribution in newly synthesized acids and presence of cell wall in the system. Biochim. Biophys. Acta 1045:58–65. 7a.Lee, R. E., K. Mikusova, P. J. Brennan, and G. S. Besra. J. Am. Chem. Soc., in press. 8. Lim, R., and S. S. Cohen. 1966. D-Phosphoarabinoisomerase and D-ribulokinase in Escherichia coli. J. Biol. Chem. 241:4304–4315. 9. McNeil, M., and P. J. Brennan. 1991. Structure, function, and biogenesis of the cell envelope of mycobacteria in relation to bacterial physiology, pathogenesis, and drug resistance; some thoughts and possibilities arising from recent structural information. Res. Microbiol. 142:451–463. 10. McNeil, M., M. Daffe, and P. J. Brennan. 1990. Evidence for the nature of the link between the arabinogalactan and peptidoglycan components of mycobacterial cell walls. J. Biol. Chem. 265:18200–18206. 11. McNeil, M., M. Daffe, and P. J. Brennan. 1991. Location of the mycolyl ester substituent in the cell walls of mycobacteria. J. Biol. Chem. 266:13217–13223. 12. Shibaev, V. N. 1986. Biosynthesis of bacterial polysaccharide chains composed of repeating units. Adv. Carbohydr. Chem. Biochem. 44:277–339. 13. Singh, S., and S. E. Hogan. 1994. Isolation and characterization of sugar nucleotides from Mycobacterium smegmatis. Microbios 77:217–222. 14. Takayama, K., and J. O. Kilburn. 1989. Inhibition of synthesis of arabinogalactan by ethambutol in Mycobacterium smegmatis. Antimicrob. Agents Chemother. 33:1493–1499. 15. Volk, W. A. 1960. Purification and properties of phosphoarabinoisomerase from Propionibacterium pentosaceum. J. Biol. Chem. 235:1550–1553. 16. Volk, W. A. 1966. D-Arabinose 5-phosphate isomerase. Methods Enzymol. 9:585–588. 17. Wojtkiewicz, B., R. Szmidzinski, A. Jezierska, and C. Cocito. 1988. Identification of a salvage pathway for D-arabinose in Mycobacterium smegmatis. Eur. J. Biochem. 172:191–203. 18. Wolucka, B. A. 1995. Lipid intermediates for arabinogalactan/arabinomannan biosynthesis in mycobacteria. J. Cell. Biochem. Abstr. Suppl. 19B:71. (Abstract.) 19. Wolucka, B. A., M. R. McNeil, E. de Hoffmann, T. Chojnacki, and P. J. Brennan. 1994. Recognition of the lipid intermediate for arabinogalactan/ arabinomannan biosynthesis and its relation to the mode of action of ethambutol on mycobacteria. J. Biol. Chem. 269:23328–23335. 20. Wright, R. S., and H. G. Khorana. 1958. Phosphorylated sugars. V. Syntheses of arabinofuranose and arabinopyranose 1-phosphates. J. Am. Chem. Soc. 80:1994–1998. 21. Xin, Y., R. Lee, K. Mikusova, G. S. Besra, M. McNeil, and P. J. Brennan. Unpublished data. 22. York, W. S., A. G. Darvill, M. McNeil, T. T. Stevenson, and P. Albersheim. 1986. Isolation and characterization of plant cell walls and cell wall components. Methods Enzymol. 118:3–40.
