Polyprenylphosphate-pentoses in mycobacteria are synthesized from 5-phosphoribose pyrophosphate.

Polyprenylphosphate-arabinose (in which the polyprenyl unit is found both as decaprenyl and octahydroheptaprenyl) is a donor of mycobacterial cell wall arabinosyl residues. Because of this important role, its biosynthetic pathway, and that of the related lipid, polyprenylphosphate-D-ribose, was investigated. Surprisingly, phosphoribose pyrophosphate was shown to be a key intermediate on the pathway to both polyprenylphosphate-D-pentoses. Thus, incubation of 5-phospho-D-[14C]ribose pyrophosphate with membranes prepared from Mycobacterium smegmatis resulted in the presence of organic-soluble radioactivity that was shown to be, in part, polyprenylphosphate-[14C]arabinose and polyprenylphosphate-[14C]ribose. Two additional intermediates, polyprenylphosphate-5-phospho[14C]ribose and polyprenylphosphate-5-phospho[14C]arabinose, were identified. Further experiments showed that the mature polyprenylphosphate-ribose is formed from phosphoribose pyrophosphate via a two-step pathway involving a transferase to form polyprenylphosphate-5-phosphoribose and then a phosphatase to form the final polyprenylphosphateribose. Polyprenylphosphate-arabinose is formed by a similar pathway with an additional step being the epimerization at C-2 of the ribosyl residue. This epimerization occurs at either the level of phosphoribose pyrophosphate or at the level of polyprenylphosphate-5-phosphoribose.

The mycobacterial cell wall core consists of a highly impermeable layer of the unique 70 -90 carbon mycolic acids covalently attached to an inner peptidoglycan layer by way of the connecting polysaccharide, arabinogalactan. Arabinogalactan consists of three regions: the linker region (1), which is connected to the peptidoglycan, a galactan directly attached to the linker (2), and an arabinan (2), which is directly attached to the galactan. The mycolic acids are attached at the non-reducing end (3) of the arabinan. Since the arabinan is fundamental to the structural integrity of the cell wall, its biosynthesis has recently been studied (4 -9) with an ultimate aim of developing new tuberculosis drugs targeted at one or more of the arabinose biosynthetic enzymes.
A major breakthrough (4) in these studies was the isolation of polyprenylphosphate-arabinose in the form of decaprenylphosphate-arabinose and octahydroheptaprenylphosphatearabinose. Subsequently, it was demonstrated that radioactive decaprenylphosphate-arabinose made chemically (5), or a mixture of radioactive decaprenylphosphate-arabinose and octahydroheptaprenylphosphate-arabinose, isolated from cultures of Mycobacterium smegmatis, 1 could function as arabinosyl donors in the presence of M. smegmatis enzymes to form polymeric arabinan. Hence, determining the pathway of the biosynthesis of polyprenylphosphate-arabinose itself becomes important for the ultimate goal of developing new drugs against mycobacteria.
In a related area, mycobacteria have been shown (9) to synthesize polyprenylphosphate-ribose. This compound has been characterized as decaprenylphosphate-ribose (9), although it is likely to exist as octahydroheptaprenylphosphate-ribose as well. The function of this glycolipid is not yet clear. It does not appear to be a biosynthetic precursor of polyprenylphosphatearabinose because radioactive polyprenylphosphate-ribose is not converted by mycobacterial enzymes to polyprenylphosphate-arabinose (9). Polyprenylphosphate-ribose may function as a ribosyl donor (9) since mycobacteria have been shown to ribosylate certain antibiotics (10).
Polyprenylphosphate sugars are generally synthesized by the transfer of a glycosyl residue from a sugar nucleotide to the phosphate moiety of a polyprenylphosphate (11). Thus, in mycobacteria, the polyprenylphosphate-mannoses (12)(13)(14) are formed by GDP-Man reacting, in a reversible fashion (13), with polyprenylphosphate to yield polyprenylphosphate-mannose plus GDP. If a similar pathway is used for polyprenylphosphate-arabinose synthesis, the sugar nucleotide of arabinofuranose needs to be clearly identified. One report has suggested the possibility of GDP-Ara (15) in mycobacteria, and a second report has suggested the possibility of UDP-Ara (6) in mycobacteria. However, the status of such arabinosyl nucleotides in mycobacterial arabinan biosynthesis remain to be clearly established.
Experiments attempting to convert radioactive polyprenylphosphate-arabinose into a sugar nucleotide of arabinose by incubating polyprenylphosphate-[ 14 C]arabinose and a variety of nucleotide diphosphates (including UDP and GDP) with enzymatically active membranes prepared from M. smegmatis were unsuccessful, 2 although such experiments succeeded with polyprenylphosphate-mannose (13). 2 Because of this result, we sought to identify an alternative pathway used by mycobacteria to synthesize polyprenylphosphate-arabinose. Recent labeling experiments using [ 14 C]glucose and cultures of M. smegmatis have shown that the carbon atoms of arabinose present in cell wall arabinan are formed by the non-oxidative pathway of the pentose shunt (7). Additionally, the obvious route from the pentose shunt to the arabinose carbon skeleton via arabinosephosphate isomerase (the enzyme converting ribulose-5phosphate to arabinose-5-phosphate) was shown not to occur (7). We therefore considered other possible routes from the pentose shunt to an activated arabinose phosphate. We report herein the results of studies using p[ 14 C]Rpp 3 as a biosynthetic precursor of polyprenylphosphate-arabinose and polyprenylphosphate-ribose.

EXPERIMENTAL PROCEDURES
Enzymes and Biochemicals-All enzymes, biochemicals, and growth media were obtained from Sigma unless otherwise noted. Radioactive D-[U- 14 C]glucose (Ͼ100 mCi/mmol) was obtained from American Radiolabeled Chemicals, Inc. (St. Louis, MO).
Preparation of Phospho[ 14 C]ribosyl Pyrophosphate-Typically, 100 Ci (100 mCi/mmol) of uniformly labeled D-[ 14 C]glucose (1 mol) was dried in a tube, and 800 l of a buffer containing 50 mM HEPES, 2 mM MgCl 2 , and 0.5 mM MnCl 2 , at a pH of 7.6, were then added. 10 units of hexokinase, 1 mol of ATP, and 4 mol of ␤-NADP were then added, and the sample was incubated at room temperature for 2 min. Then, sequentially, 10 units of glucose-6-phosphate dehydrogenase, 2 units of 6-phosphogluconate dehydrogenase, 10 units of phosphoriboisomerase, and 1 unit of phosphoribosyl pyrophosphate synthetase were added. The reaction was incubated at 37°C for 30 min. Then, a total of 1.5 mol of additional ATP was slowly added over 60 min. This resulted in p[ 14 C]Rpp with a radiopurity of over 85% as analyzed by HPLC. The protein was removed by centrifugation through a 10-kDa cutoff microcon "microconcentrator" (Amicon, Inc.). For most experiments the p[ 14 C]Rpp was further purified by preparative Dionex chromatography. The entire sample (ϳ1 ml) was injected onto a "semi-preparative" Dionex Carbopak PA1 column (9 ϫ 250 cm) and eluted with a linear gradient beginning with 0.05 M sodium acetate in water and increasing to 1 M sodium acetate in water over 30 min. The flow rate was 3 ml/min, 1.5-ml fractions were collected, and a small aliquot of each fraction was counted. The p[ 14 C]Rpp elutes at about 27 min, and fractions containing p[ 14 C]Rpp were combined. Attempts to desalt HPLC-purified p[ 14 C]Rpp resulted in substantial decomposition of the p[ 14 C]Rpp. However, since the sodium acetate did not have a deleterious effect on the M. smegmatis enzymatic activities described below, the p[ 14 C]Rpp was used without removing the salt. The purified material was frozen in aliquots and typically contained 10,000 cpm/l with a sodium acetate concentration of approximately 0.7 M.
M. smegmatis Membrane Enzyme Preparation-M. smegmatis (strain MC 2 155) was grown to mid-log in nutrient broth, harvested by centrifugation, and transferred into buffer A, which consists of 50 mM MOPS, 5 mM MgCl 2 , and 0.5 mM dithiothreitol, at a pH of 7.8. The cells, suspended in buffer A, were disrupted by sonication at 4°C with a 4710 series Cole Parmer ultrasonic homogenizer equipped with a microtip. The cell homogenate was then centrifuged at 12,000 ϫ g for 10 min, and the resulting supernatant was measured for protein concentration using the Bio-Rad Protein Assay (Bio-Rad) and shown to be typically around 10 mg/ml. The supernatant was then ultracentrifuged at 100,000 ϫ g for 1 h, and the resulting membrane-enriched pellet was then removed and homogenized in a small amount of buffer A. Its protein concentration was adjusted to approximately 4 mg/ml with buffer A, unless noted otherwise. This was then used as the "M.  (Fig. 1A, lane 1). A second reaction was done in the same fashion except that 0.8 mol of ATP and 0.8 mol of NADP were included (Fig. 1A, lane 2). After the incubation, the reactions were quenched and distributed into an organic layer and an aqueous layer by the addition 1.33 volumes of methanol and 2.67 volumes of chloroform and briefly centrifuged at 2,000 ϫ g. The organic (bottom) layer contained the newly formed polyprenylphosphate-[ 14 C]pentoses.
TLC Analysis of Lipid Products and Autoradiography-Samples were loaded on an aluminum-backed silica gel (60 F 254 ; E. Merck, Darmstaddt, Germany) TLC plate (10 ϫ 20 cm) and run in the solvent CHCl 3 :CH 3 OH:7 M NH 4 OH in H 2 O 65:25:4. Autoradiography was carried out by exposing X-OMAT AR film (Kodak) to the TLC overnight at Ϫ70°C. Samples were removed from the TLC by scraping the band into a test tube and extracting with the same solvent used to develop the TLC.
HPLC Analyses for Monosaccharides, Pentose-5-phosphates, and pRpp-Monosaccharide analysis was as described using Dionex HPLC (7) except that when the released sugars were suspected to be 5-phosphorylated, the dry hydrolysate was then taken up in 90 l of buffer A containing 1 unit of alkaline phosphatase to remove the phosphate groups. Radioactivity eluting from the Dionex HPLC was either monitored by the collection of 30 1-min fractions and counting them on a liquid scintillation system or via direct detection of the column effluent with a ␤-RAM on-line radioactivity HPLC detector scintillation system (Inus Systems, Tampa, FL).
For the detection of pentose-5-phosphates, the samples were hydrolyzed for 30 min with 0.1 M trifluoroacetic acid at 120°C. The 5-phosphate group is stable to the acid in these conditions, whereas the phospho-linkage at carbon 1 is not. The samples were then injected onto the Dionex-PA1 column and eluted with a gradient of 0.1 M sodium acetate in 2 mM NaOH to 0.8 M sodium acetate in 2 mM NaOH over 50 min.
For the detection of pRpp, the sample was injected onto the PA-1 column and eluted with a gradient from 50 mM to 1 M sodium acetate in water over 50 min. Note that for elution of pRpp, no base is present in the eluent.

Treatment of p[ 14 C]Rpp with Membranes Prepared from M. smegmatis in the Presence of Various
Additives-Six treatments are reported (Fig. 2). All treatments were done in a total volume of 140 l of buffer A and contained 400,000 cpm of p[ 14 C]Rpp (non-HPLC purified), 14 g of ethambutol (to minimize polymerization of any polyprenylphosphatearabinose formed), and 100 l of M. smegmatis membranes (13 mg/ml protein). For treatment 1, the membrane preparation was boiled 10 min before use; for treatment 2, there were no additional reagents; and for treatment 3, 25 g of decaprenylphosphate were added. For treatment 4, the membrane preparation was boiled for 10 min before use, and 1.4 mg of CHAPS and 25 g of decaprenylphosphate were added. For treatment 5, 1.4 mg of CHAPS were added, and for treatment 6, 1.4 mg of CHAPS and 25 g of decaprenylphosphate were added. When used, decaprenylphosphate was dried in the tube beforehand and solubilized by a brief pan sonication in the MOPS buffer system. The tubes were incubated and extracted as described above and an aliquot of the organic phases was counted (Fig. 2).

Treatment of p[ 14 C]Rpp with M. smegmatis Membranes in the Presence of Possible
Inhibitors-Four treatments were done. All treatments were done in a total volume of 500 l of buffer A and contained 470,000 cpm of HPLC-purified p[ 14 C]Rpp, 50 g of ethambutol, 4 mol of ATP, 4 mol of NADP, 5 mg (1%) CHAPS, 20 g of decaprenylphosphate, 70 mol of sodium acetate (introduced from the HPLC-purified p[ 14 C]Rpp), and 350 l of M. smegmatis membranes (3.5 mg/ml protein). The four treatments varied in that 4 mM Na 2 WO 4 , 4 mM NH 4 VO 3 , 5 mM NaF, or no potential inhibitors were present. After incubation for 30 min at 37°C, the samples were extracted, and the organic layer was analyzed by TLC (Fig. 7).
Preparation of Polyprenylphosphate-pentose Standards-Decaprenylphosphate-[ 14 C]arabinose was chemically synthesized as described previously (5 it was expected to contain both the arabinose and ribose sugar components. This material was prepared, purified by TLC, and removed from the TLC plate as described above (see Fig. 1A, R f 0.14). The preparation is thus a mixture of polyprenylphosphate-5-phospho[ 14 C]arabinose and decaprenylphosphate-5phospho[ 14 C]ribose. The second preparation of polyprenylphosphate-5pentoses was made by incubating p[ 14 C]Rpp with M. smegmatis membranes as described above but included 4 mM Na 2 WO 4 , 100 g/ml ethambutol, and 1% CHAPS. It was expected to only have ribo pentosyl units because CHAPS inhibits formation of the arabinosyl lipids. TLC analysis of this second material (see Fig. 6, lane 3) showed that the putative polyprenylphosphate-5-phosphoribose was the only radioactive component, and due to concerns of its breaking down during TLC purification, it was used directly. An aliquot (10,000 cpm) of the putative polyprenylphosphate-5-phospho[ 14 C]ribose was treated directly with alkaline phosphatase. In this instance, the sample was resuspended in 100 l of buffer A containing 1% CHAPS by pan sonication. An additional 500 l of buffer A was added, which contained 2 units of alkaline phosphatase, and the sample was incubated at 37°C for 30 min. The sample was subjected to organic extraction, and the organic layer was analyzed by TLC (Fig. 6, lane 2).
Conversion of Polyprenylphosphate-5-phosphoribose to Polyprenylphosphate-ribose by M. smegmatis Membranes-An aliquot (10,000 cpm) of the polyprenylphosphate-5-phospho[ 14 C]ribose was resuspended in buffer A (200 l). Then, 400 l of M. smegmatis active membrane preparation (3.5 mg/ml protein) were added, and the sample was incubated at 37°C for 30 min. After extraction, the organic layer was analyzed by TLC (Fig. 6, lane 1 4 were resuspended in buffer A (50 l). One of the aliquots received PP i so that the final concentration was 4 mM, a second 2 mM, and the third and fourth received no PP i . To tubes 1, 2, and 3, 92 l of active M. smegmatis membrane preparation (6.3 mg/ml protein) were added; to tube 4, 92 l of boiled membrane preparation were added. After incubation for 30 min at 37°C, the samples were extracted, and the aqueous and organic layers were counted.
A large scale experiment was done by suspending 200,000 cpm of the polyprenylphosphate-5-phospho[ 14 C]ribose into 100 l of buffer a containing 1% CHAPS. One-and two-tenths mol of PP i in 50 l of buffer were added along with 150 l of active membrane. After incubation for 30 min at 37°C, the material was extracted, and the aqueous phase was analyzed by Dionex HPLC for p[ 14 C]Rpp (Fig. 8). Approximately 1% of the starting radioactivity was converted into organic soluble material as compared to less than 0.1% in an enzyme boiled control. The organic soluble material from two reactions using active membranes was run on TLC, and the resulting autoradiogram is presented in Fig. 1A. The materials with R f s of 0.43 and 0.50 were removed separately from the TLC, hydrolyzed, and analyzed by HPLC (Fig. 1, B and C) for [ 14 C]sugars. The material with a R f of 0.5 was mostly arabinose, and the material with R f of 0.43 was about an equal mixture of arabinose and ribose. Previous studies (4,9) have shown that of the polyprenylphosphate-pentoses, decaprenylphosphate-arabinose runs the fastest on TLC, with octahydroheptaprenylphosphate-arabinose and decaprenylphosphate-ribose migrating nearly together (9) and slightly slower than decaprenylphosphate-arabinose. This information, combined with the sugar analyses ( Fig. 1, B and C) allowed for the material with R f of 0.50 to be identified as decaprenylphosphate-[ 14 C]arabinose and the material with R f of 0.43 to be identified as a mixture of decaprenylphosphate-[ 14 C]ribose and octahydroheptaprenylphosphate-[ 14 C]arabinose (see below for further substantiation of these identities).

The Effect of Various Additives on the Conversion of p[ 14 C]Rpp into Organic Soluble Material-Since the conversion of p[ 14 C]
Rpp into polyprenylphosphate lipid catalyzed by the M. smegmatis membranes was low, the effect of various additives on this conversion was investigated (Fig. 2). Most striking was the dramatic stimulation of activity by the detergent CHAPS (Fig. 2, treatment 5). Also important was the stimulation of activity by decaprenylphosphate, which could be seen both in presence and absence of the detergent (Fig. 2, treatments 6 and 3). TLC analysis of the material treated with both CHAPS and decaprenylphosphate gave a profile similar to that shown in Fig. 1 except that sugar analysis of the faster migrating material yielded only ribose. These results led to the conclusion that the conversion of pRpp to polyprenylphosphatepentose was strongly stimulated by the detergent, but only polyprenylphosphate-ribose was formed. Clearly, one of the enzymes involved in the arabinose lipid formation fails to act in the presence of CHAPS.
Evidence for the Nature of the Lipid Group Present on the Glycolipids Formed from p[ 14 C]Rpp-As noted above, the TLC migration times strongly suggested that the lipid moieties of the organic soluble glycolipid were polyprenylphosphates. To compare TLC migrations directly, preparations of the polyprenylphosphate-pentoses synthesized from p[ 14 C]Rpp were isolated, and their TLC mobility was compared with standards. Thus, a mixture of decaprenylphosphate-[ 14 C]arabinose and decaprenylphosphate [ 14 C]ribose was synthesized using membranes and p[ 14 C]Rpp in the absence of detergent. The polyprenylphosphate-pentose region of the TLC was eluted and shown by sugar analysis to be 50% arabinose and 50% ribose. This material was then compared by TLC to a chemically synthesized standard of decaprenylphosphate arabinose (Fig.  3A); the more mobile band (shown by the analysis illustrated in Fig. 1  Further conformation of the nature of the lipid and of the fact that a polyprenylphosphate was utilized as a substrate by the membrane enzymes was obtained by performing incubations in the presence and absence of pentadecaprenylphosphate (C75). It was expected that if this exogenously added lipid was used as an acceptor to form pentadecaprenylphosphate-pentoses, faster migrating bands would be evident due to the chain length of pentadecaprenylphosphate. Examination of Fig. 4 shows that this expectation was realized with a faster migrating band (R f 0.52) in the polyprenylphosphate-pentose region and a faster band (R f 0.09) in the "slowly migrating" region of the TLC.
Characterization of the Slowly Migrating (R f 0.07-0.14) 14 C-Labeled Lipids-The slowly migrating lipids (Fig. 1, R f 0.14) were suspected to be a mixture of polyprenylphosphate-5-phosphoarabinose and polyprenylphosphate-5-phosphoribose. Consistent with this hypothesis, acid hydrolysis followed by alkaline phosphatase (to remove the phosphate from the 5 position) released both arabinose and ribose from these lipids as shown in Fig. 5A. Similar treatment of the analogous glycolipid formed from p[ 14 C]Rpp in the presence of detergent yielded only [ 14 C]ribose (Fig. 5B). In a different analysis, the material at R f 0.14, which contained both arabinose and ribose (Figs. 1 and 5A), was treated with mild acid to remove the expected pentose-5-phosphates from the lipid in an intact form. The resulting sugar phosphates were analyzed by HPLC. The results (Fig. 5C) showed the presence of both ribose-5-phosphate and arabinose-5-phosphate. As anticipated, similar analysis of the slowly migrating lipid formed in the presence of detergent revealed only ribose-5-phosphate (results not presented). Finally, treatment of the slowly migrating ribose glycolipid with alkaline phosphatase converted it to mature polyprenylphosphate-ribose (Fig. 6, lane 3 before treatment and lane 2 after treatment). Therefore, the structures of the slowly migrating glycolipids were assigned to be polyprenylphosphate-5phospho[ 14 C]ribose and polyprenylphosphate-5-phospho[ 14 C] arabinose.

Evidence for the Conversion of Polyprenylphosphate-5-phosphoribose to Polyprenylphosphate-ribose as Part of the Polyprenylphosphate-pentose Biosynthetic Pathway in Mycobacteria-
The polyprenylphosphate-5-phosphopentoses, due to their structures, are very likely to be the direct biosynthetic precursors of the mature polyprenylphosphate-pentoses. Direct evidence of such was obtained for the ribo compounds by using known inhibitors of phosphatases. Thus, Na 2 WO 4 , NH 4 VO 3 , and NaF were added, in the presence of CHAPS, to the p[ 14 C]Rpp and M. smegmatis membranes. After incubation and extraction, the organic soluble materials were run on TLC as shown in Fig. 7. As is evident, the formation of mature polyprenylphosphate-[ 14 C]ribose was strongly inhibited by the tungstate (94%), partially inhibited by the vanadate (32%), and partially inhibited by the fluoride (30%). In a more direct approach, polyprenylphosphate-5-phospho[ 14 C]ribose was isolated and then re-incubated with M. smegmatis enzymes. Such treatment converted the polyprenylphosphate-5-phospho-[ 14 C]ribose to polyprenylphosphate-ribose (Fig. 6, lane 3 before re-incubation and lane 1 after re-incubation).
Polyprenylphosphate-5-phospho[ 14   radioactivity to aqueous soluble radioactivity. The aqueous soluble radioactive material produced in the presence of 4 mM PP i was analyzed by Dionex HPLC and shown to co-elute with p[ 14 C]Rpp (Fig. 8). DISCUSSION The experiments reported herein allowed for the elucidation of the biosynthetic pathway for the formation of polyprenylphosphate-ribose (Fig. 9). Thus, a 5-phosphoribosyl transferase transfers a ribose-5-phosphate unit from the pyrophosphate moiety of pRpp to the phosphate moiety of decaprenylphosphate to form decaprenylphosphate-5-phosphoribose. This transfer occurs with an inversion of configuration at the 1 position of the ribosyl unit, as pRpp is in the ␣ configuration and decaprenylphosphate-ribose is in the ␤ configuration (9). The reaction is novel but follows the same general principles as those used to form polyprenylphosphatehexoses. The difference is that with the hexoses, the leaving group is a nucleotide-pyrophosphate (11) rather than inorganic-pyrophosphate.
The second step of the reaction is the dephosphorylation of polyprenylphosphate-5-phosphoribose to form the mature polyprenylphosphate-ribose (Fig. 9). The phosphatase catalyzing this reaction is found in the membrane fraction and is strongly inhibited by tungstate (Fig. 7).
The experiments reported herein show that pRpp is a biosynthetic precursor of polyprenylphosphate-arabinose and that polyprenylphosphate-arabinose is formed via polyprenylphosphate-5-phosphoarabinose. Clearly, an epimerization at C-2 must occur given the ribo stereochemistry of pRpp and the arabino stereochemistry of polyprenylphosphate-arabinose. The evidence for the existence of polyprenylphosphate-5-phosphoarabinose (Fig. 5) suggests that pRpp might be epimerized to form 5-phosphoarabinose pyrophosphate, which then reacts with decaprenylphosphate to form polyprenylphosphate-5phosphoarabinose (path A, Fig. 9). Alternatively, the epimerization might occur at the lipid level with polyprenylphosphate-5-phosphoribose being epimerized to form polyprenylphosphate-5-phosphoarabinose (path B, Fig. 9). Of these two possibilities, we presently favor the former because of earlier results (16) consistent with the inter-conversion of pRpp and 5-phosphoarabinose pyrophosphate.
Unexpectedly, pentose sugar nucleotides are not involved in the biosynthetic pathway for polyprenylphosphate-pentose formation (Fig. 9). This finding does not preclude the formation of an arabinofuranose sugar nucleotide in a different pathway and its utilization by some mycobacterial arabinosyl transferases as a donor. However, since recent studies 1 have shown that many, if not all of the arabinosyl residues present in mycobacterial arabinan, arise from polyprenylphosphate-arabinose, the distinct possibility exists that mycobacteria synthesize their arabinofuranosyl polymer entirely without the utilization of an arabinofuranosyl sugar nucleotide.
Inhibition of the polyprenylphosphate-5-phosphoarabinose phosphatase should inhibit the formation of polyprenylphosphate-arabinose and, hence, the formation of arabinan and the cell wall. Similarly, inhibition of the enzyme that forms poly- prenylphosphate-5-phosphoarabinose, whether it is an epimerase or transferase, should inhibit mycobacterial cell wall formation. Inhibitors of these two target enzymes would likely be nontoxic given the lack of D-arabinofuranosyl residues in humans.
The most important unresolved issue concerns the point at which the D-ribo to D-arabino conversion takes place in the formation of polyprenylphosphate-arabinose. Research efforts in this direction are now proceeding.