Biosynthesis of the Mycobacterial 0-Methylglucose Lipopolysaccharide CHARACTERIZATION OF PUTATIVE INTERMEDIATES IN THE INITIATION, ELONGATION, AND TERMINATION REACTIONS*

From the 70% ethanol extract of Mycobacterium smegmatis cells, we isolated a mixture of weakly acidic oligosaccharides composed mainly of glucose and 6-0- methylglucose. The elution pattern from a Bio-Gel P-4 column suggested that the oligosaccharides were smaller than the 0-methylglucose polysaccharide (MGP) and could be biosynthetic precursors. Analysis by fast-atom-bombardment mass spectrometry re- vealed that the oligosaccharides fit into a pattern for polysaccharide synthesis based on an alternate gluco- sylation-methylation mechanism until the chain reached the composition methylglucosellglucose5gly- ceric acid, at which time 2 glucose units are added to give glucosezmethylglucose11glucose5glyceric acid. The addition of the last 2 glucoses and methylation of one of them to give mature MGP (methylglucose,-glucosesmethylglucose11glucose5glyceric acid) appar- ently occurs rapidly because the expected intermediates were not observed. Only 4 glucose units are pres- ent at the glyceric acid end of some molecules during all stages of the elongation process, and these represent precursors of a minor MGP homolog with an extra methyl group on the 8143-linked glucose unit of MGP. a-~-Glucopyranosyl-( (-400 mesh) column (2 X 190 cm) by elution with the same solvent. The other acidic fractions were treated in a similar manner.

( 1 To whom inquiries should be addressed. elongated by a sequential mannosylation-methylation reaction, during which guanosine diphosphate mannose serves as the hexose donor and S-adenosylmethionine as the methyl donor. Termination of the elongation reaction occurs when the chains reach a length with good fatty acid-binding properties (11-13 3-0-methylmannose units), and the chains end with an unmethylated mannose because the K,,, for the 3-0methyltransferase is much higher than that of the (~( 1 4 )mannosyltransferase after the chains reach such a length and form lipid complexes. These enzymic studies are supported by the fact that the expected intermediates (MeMan,-OMe' and Man-MeMan,-OMe, where x = 4-10) can be isolated from a 70% ethanol extract of mycobacterial cells (3).
The search for precursors of the methylglucose lipopolysaccharide has, until now, been less fruitful because the expected intermediates accumulate in very small amounts and because cell extracts contain other unrelated acidic partially methylated glucans. In the present report, however, we describe the isolation of weakly acidic and partially methylated glucans of the size predicted for the putative precursors of the methylglucose polysaccharide, and analysis by fast-atom-bombardment mass spectrometry reveals that the oligosaccharides fit a pattern of sequential glucosylation-methylation for biosynthesis of the 6-0-methylglucose-containing portion of the polymer. The data also support a novel mechanism for termination of the elongation reaction.

EXPERIMENTAL PROCEDURES
Materials and Methods-Sources for most of the materials used in this study are listed in earlier publications (13,14). Affi-Gel601, for binding sugars with cis-diols, was from Bio-Rad. Mycobacterium smegmatis ATCC 356 was grown in a 200-liter fermentor on a glycerol medium ( E ) , and the cells were harvested with a large steam-driven Sharples centrifuge to give about 2 kg of wet cell paste. The cells were stored frozen until required. The methodologies for gas, thin layer, and ion exchange chromatography, for gel filtration, for mass spectrometry, and for colorimetric determination of carbohydrate have been described (13).
Isolation of Putative MGP Precursors-M. smegmatis cell paste (500 g) was stirred in acetone (2 liters) at 20-23 'C for 18 h, and the insoluble cell debris was collected on a Buchner funnel with glass filter paper, washed with acetone, and dried in air. The acetonewashed cells were extracted twice for 2 h each time with refluxing 70% ethanol (2 liters), and the combined ethanolic filtrates were concentrated to a small volume and saponified with 1 N NaOH (100 ml) at 100 "C for 8 h. After the cooled solution was neutralized with The abbreviations used are: OMe, methoxy; MGP, 6-0-methylglucose polysaccharide; AGMGP, amylase-and glucoamylase-digested MGP that removes 4 hexose units; Me, methyl group; Gla, glyceric acid; Glc, glucose; Man, mannose; FAB, fast atom bombardment; FABMS, fast-atom-bombardment mass spectrometry; Hex, hexose.

4580
Dowex AG 50W-X8 resin, it was extracted with diethyl ether (200 ml) twice to remove any lipids and concentrated to dryness. To remove strongly acidic material, the residue was dissolved in water (50 ml) and applied to a column of DEAE-Sephadex A-25 (HCO;) (150 ml), which was washed with 40 mM ammonium bicarbonate (500 ml). The effluent which contained neutral and weakly acidic materials was lyophilized to remove most of the salt, the residue was treated with 1 N NaOH (25 ml) at room temperature for 5 h to saponify any glyceric acid lactone that may have formed during lyophilization, and the cations were again removed with Dowex AG 50W-X8. The solution was applied to a column of DEAE-Sephadex A-25 (HC05) (150 ml), which was washed with water to remove a neutral fraction and then eluted stepwise with 10,20, and 40 mM ammonium bicarbonate (400 ml each) to yield acidic fractions. The yield of carbohydrate as glucose was 18.1 mmol (neutral fraction), 3.0 mmol (10 mM acidic fraction), 0.43 mmol (20 mM acidic fraction), and 0.69 mmol (40 mM acidic fraction). Following elution of the first DEAE column with 40 mM salt, the strongly acidic fraction was eluted with 80 mM salt to give 0.43 mmol of carbohydrate.
The 10 mM acidic fraction (200 mg of carbohydrate) was dissolved in 2 ml of 0.25 M ammonium acetate, pH 8.8, and the solution was passed through a 5-ml column of Affi-Gel 601 to remove mannosecontaining materials. The column was washed with three 5-ml portions of the same buffer to recover unbound material, and the effluent was concentrated and lyophilized several times after addition of water to remove the salt. The residue was dissolved in 2 ml of 0.1 M acetic acid and fractionated on a Bio-Gel P-4 (-400 mesh) column (2 X 190 cm) by elution with the same solvent. The other acidic fractions were treated in a similar manner.

RESULTS
Isolation of MGP Precursors-The 70% ethanol extract of acetone-dried M. smegmatis cells contains many oligosaccharide components, some of which are acylated. To remove the acyl groups completely, the material had to be treated in hot alkaline solution for several hours. Because MGP and its precursors are weak acids, it was important to decationize the resulting solution completely to ensure efficient binding of the oligosaccharides to the DEAE-Sephadex column. The unbound material yielded mostly mannose, arabinose, glucose, and 3-0-methylmannose on acid hydrolysis, whereas oligosaccharides rich in 6-0-methylglucose were eluted with 10 mM salt. The 20 mM salt fraction also gave some 6-0-methylglucose on hydrolysis, but the materials eluted at higher salt were composed mainly of glucose, mannose, arabinose, and other sugars.
Because arabinose and mannose related to a known acidic arabinomannan (16) contaminated hydrolysates of most fractions, the MGP precursors were purified by passing the solutions through a phenyl boronate column, which bound most of the former but not the latter. The unbound material from the 10 mM salt elution was then fractionated according to size on a Bio-Gel P-4 column. The vast majority of material was eluted in the position of mature MGP, which peak was separated into 3 fractions, while the trailing end of the MGP peak was separated arbitrarily into several fractions as indicated on Fig. 1. The monosaccharide compositions of the fractions revealed that many of the samples were still contaminated with small amounts of sugar-containing materials unrelated to MGP, but these did not interfere with the mass spectral analysis. Estimates of the degree of polymerization of oligosaccharides in these fractions must take into account the facts that each methyl ether group contributes the equivalent of 1.6 hexose units (17) and that the glyceric acid group contributes about 2 hexoses to the apparent size because of its charge.
The material isolated from the DEAE column by elution with 20 mM salt was treated in a similar manner, and the distribution of carbohydrate in the Bio-Gel P-4 column effluent is shown in Fig. 2. The large peak in the void volume is related to the acidic arabinomannan (16) mentioned above, whereas the other two larger peaks between fractions 180 and 200 are related to a glycolipid (13) and a phospholipid (18, 19) described previously.
The 40 mM salt eluate of the DEAE-Sephadex column gave a single carbohydrate-containing peak by gel filtration on a Bio-Gel P-4 column that appeared at the position of a trisaccharide (not shown). This material proved to be identical to fraction 15 in Fig. 2, and it was characterized as glucosylglyceric acid (20,21) (see below). Elution of the DEAE column with 80 mM salt gave more carbohydrate-containing material that was rich in glucose, arabinose, mannose, and phosphate, which suggested that this fraction contained trehalose phosphate (10) and glucolipid (13) and also was related to the acidic arabinomannan (16). Preliminary analyses confirmed this conclusion, but further characterization was not attempted.
Analysis of the Acidic MGP Precursor Fraction Eluted by 10 mM Salt-The majority of carbohydrate in the putative precursor fraction, eluted from DEAE-Sephadex with 10 mM salt and fractionated by gel filtration on a Bio-Gel P-4 column (Fig. l ) , appeared at a position characteristic of MGP, but a small amount trailed after the main peak to a position at which an eicosasaccharide (Glczo) would appear. From other work (17), we expect an oligosaccharide composed of 5 0methylglucoses, 5 glucoses, and a glyceric acid unit to be eluted at the Glc2o position, whereas MGP is eluted near the void volume of a Bio-Gel P-4 column.
The fractions from the Bio-Gel P-4 column were combined arbitrarily as indicated on the figure and the FAB mass spectrum of each was determined in the positive mode. The major ions in the spectra of all 8 fractions (Table I)  The FAB mass spectrum of MGP (19) previously demonstrated that about 20% of the molecules have an extra methyl group, which must be near the glyceric acid end of the chain because it is also present in AGMGP (22). This minor homolog is enriched in fraction 1 of Fig. 1, and we have used this material to determine the exact location of the methyl group. The sample was digested with Rhizopus amylase to remove most of the 6-0-methylglucose chain (14), and the acidic limit product was isolated by ion exchange and fractionated on a Bio-Gel P-4 column (Fig. 3). Peaks A and C correspond to fractions A-1 (MeGlc3Glc6Gla) and A-2 (MeGlc4Glc5Gla) from a previous study (14) and are derived from the major MGP component. Peaks B and D are displaced on the gel filtration pattern by an amount suggesting that they differ from peaks A and C, respectively, by one methyl group (apparent volume = 1.6 hexoses/methyl). The compositions (MeGlc4Glc4Gla and MeGlcSGlc4Gla) confirm this and demonstrate that the extra methyl group is on position 6 of one of the 5 glucose units in A-1 and A-2 (Table 11).
An attempt was made to determine the location of the extra methyl group from the anomeric proton NMR spectra (Table  111). The signals for the 4 glucoses nearest the glyceric acid (A, B, C, and D in Fig. 4) are shifted no more than 0.003 ppm on addition of the methyl group, whereas 3 of the other 4 hexoses are significantly affected. The anomeric proton of the a(l-~3)-linked glucose is shifted upfield 0.024-0.027 ppm, whereas the methylglucose to which it is attached is slightly shielded and the other two 6-0-methylglucoses are deshielded about 0.025 ppm. We found previously that methylation at position 6 of glucose has a relatively small effect on the anomeric proton of that hexose, but it can affect the shifts of neighboring glucoses (14). Therefore, we concluded that the NMR spectra do not allow a definitive assignment.
Convincing evidence that the extra methyl group was on glucose D (Fig. 4)  Fractions 5-8 from Fig. 1 show an interesting and suggestive pattern in which the ions correspond to molecules that contain, alternately, 4 and 5 glucoses along with varying numbers of 0-methylglucoses (Table I). The 5th glucose in each instance is represented as occurring at the glyceric acid end of the chain (MeGlc,Glc5Gla). Support for this conclusion was obtained by analysis of the products formed by Rhizopus amylase digestion of the oligosaccharide mixture. Such digestion of mature MGP yields MeGlc3Glc5Gla and Me-Glc,GlclGla (see above) (14), and digestion of the combined precursor fractions 5-8 gave ions on FABMS at m/z 1443 and 1457, which correspond to these same fragments.
Fraction 4 from Fig. 1 gave a complex assortment of ions that extends the series found in fractions 5-8 up to the homologs with 11 and 12 methylglucoses, at which point ions are observed corresponding to the additions of 1 and 2 glucose units without further methylation. This is consistent with a biosynthetic pathway in which, following construction of the   ' Refer to Fig. 4. Fig. 3.

Assignments arbitrary. Other assignments based on comparison
Overlapping signals prevented measuring coupling constants.
In hexose units on a Bio-Gel P-4 column calibrated with maltooligosaccharides. The numbers in parentheses are the apparent sizes calculated assuming that each glucose has a value of 1, each 6-0methylglucose has a value of 2.6, and the glyceric acid has a value of 2.
Determined by gas chromatography on an OV-225 column, by which method all monomethylglucitol pentaacetates have characteristic retention times. NO peak for a dimethyl derivative was observed.  (Table IV). To compare the ions of the two tables, one need only add or subtract 24 mass units as appropriate. Fractions 1-3 of Fig. 2 gave ions for several intermediates that overlap with those in fractions 7 and 8 of Fig. 1, whereas fractions 4-12 of Fig. 2 gave the expected ions extending the homologous series down to MeGlclGlcsGla (m/z 1091). Two ions were seen (m/z 1957 and 21331, with intensities about 10% of the major ions in the spectra, that correspond to intermediates with an extra unmethylated glucose we have placed at the nonreducing end and are presumed to be the methyl acceptors for synthesis of the next higher homologs (m/z 1971 and 2147). Another ion (m/z 943) corresponded to MeGlczGlcsGla and could be a precursor of the more highly methylated form of MGP (m/z 3551). An ion at m/z 735 and one at m/z 657 reveal that fractions 11 and 12 are contaminated with non-MGP fragments related to a glycolipid (13) and a phospholipid (18) described previously.
The assignments of structures to the ions listed in Table  IV are supported by the FAB mass spectrum of the Rhizopus amylase limit digest of the combined fractions 1-5. The acidic fraction from the digest gave ions for MeGlc,Glc,Gla (where x = 4-6) and MeGlc,Glc6Gla (where x = 3 4 , and no ion for a compound with 6 glucoses was observed, as expected if the 6th glucose in fractions 1 and 3 was in a terminal position where it would be removed by amylase.
Finally, fractions 14 and 15 of Fig. 2 gave single ions  corresponding, respectively, to Glc2Gla and GlclGla, that could represent the first two intermediates in MGP synthesis. The assignments of these two ions are consistent with their elution positions on the Bio-Gel P-4 column because the glyceric acid moiety contributes the equivalent of 2 hexoses to the apparent volume of a glycoside. It is notable that ions for intermediates between Glc2Gla and MeGlclGlcaGla were not observed. By gel filtration of the material eluted from DEAE-Sephadex with 40 mM salt, a single carbohydratecontaining peak was obtained that corresponded to GlclGla and gave a single ion at mlz 267. This sample was character-  IsC chemical shift a The first value is for the ionized form and the second for the protonated form of the glyceric acid moiety.

Fractions and ions observed in negative mode [M -H]"
3.8-3.9 ppm. We assign the set at 64.94 and 5.02 to a -~glucopyranosyl-(1~)-a-D-glucopyranosyl-(l~2)-D-glyceric acid, an expected MGP precursor, whereas the other set at 65.39 and 4.95 suggests the presence of a related compound with the structure a-D-glucopyranosy~-(1~)-a-D-glucopyranosyl-(1+2)-D-glyceric acid. The assignment for the first compound is supported by the chemical shifts we observed for a diglucosyl-glyceric acid obtained by partial acid hydrolysis of MGP (64.94 and 5.02). We are unable to account for the second diglucosyl-glyceric acid, but it could be a precursor of some other unreported oligosaccharide.

DISCUSSION
The biosynthesis of the 0-methylglucose lipopolysaccharide is interesting because the polymer has a precisely defined length and sequence, which is unusual for polysaccharides of this size. Since the structure of polysaccharides is determined solely by enzyme specificity, this implies that the a(1-A)glucosyltransferase(s) can recognize when the 15th in a chain of similarly linked sugars has been added, and the 6-0methyltransferase(s) can sense when the 11th glucose unit in the otherwise monotonous a(lA)-linked chain has been methylated. Second, the polysaccharide chain is acylated in 8 specific positions (23) by 5 different kinds of acyl groups (24), which indicates the existence of several highly specific acyltransferases (25). In previous studies in which maltooligosaccharides, a(14)-linked glucans, were used as exogenous acceptors for the putative methyl-and acyltransferases, it was found that the acceptor ability in both reactions was dependent on partial acetylation (17) implying that methylation and acylation may proceed together.
The present report documents the existence in M. smegmatis cell extracts of smaller oligosaccharides related in their structures to the methylglucose polysaccharide and of such a composition as to support a mechanism for elongation of the chain based on sequential glucosylation-methylation. Weakly acidic oligosacharides with the composition MeGlc,Glc5Gla, where x = 1-11, were detected by mass spectrometry, along with a parallel series with the composition MeGlc,Glc,Gla, where x = 2-12. These structures reveal that the precursors have 4 or 5 glucose units at the glyceric acid end of the chain in agreement with the fact that mature MGP is a mixture of isomers that possesses 4 or 5 glucoses in this region.
The mass spectral data further show that the addition of glucose units to the oligosaccharide MeGlcllGlc5Gla then occurs without further methylation to yield the presumed intermediate GlczMeGlcllGlc5Gla. Conversion of this oligosaccharide to mature MGP requires the addition of two more glucoses to the terminal Glcz unit and methylation of the last one. The expected intermediates for these steps were not observed, however, which suggests that the reactions may occur rapidly and in concert to give mature MGP. We expect that the oligosaccharide with the presumed structure GlczMeGlcllGlc5Gla will associate readily with long-chain fatty acid to form an inclusion complex in which the polymer assumes a tightly coiled conformation (9), and it is possible that such a complex is presented to the transferases that catalyze the termination reactions. Such postulated coiling might inhibit further 6-0-methylation and promote 3-0methylation. If such a mechanism is involved, it would parallel in some ways that which appears to regulate termination of methylmannose polysaccharide elongation (12).
In Fig. 5, we have selected representative structures detected by FABMS and have ordered them into a plausible pathway for biosynthesis of both forms of MGP. It is apparent that the extra methyl group in the minor homolog is introduced at an early stage in the process, but we cannot confute that it might also be added to some molecules at later times. That most of the methylated intermediates are terminated by 0-methylhexose suggests that the 0-methyltransferase must have a lower K,,, for the acceptor than the glucosyltransferase. Two intermediates possessing 6 unmethylated glucose units were detected, however, and our structural analysis indicates that the sixth glucose is at the nonreducing end of the chain where it could serve as the methyl acceptor in the next step. Support for the conclusion that the MGP precursors consist of homologs with 4 or 5 glucoses at the glyceric acid end of the chain came from analysis of the limit product resulting from digestion of the oligosaccharides with a mold amylase that catalyzes hydrolysis of 6-O-methyl-a-~-glucosyl-( 1 4 ) linkages (14). The product had the composition MeGlc,Glc4Gla, MeGlc,Glc4Gla, MeGlc3Glc5Gla, and MeGlc4Glc5Gla, which also are obtained when a mixture of mature MGP-I and MGP-I1 is digested in a similar manner.
Our analysis of the MGP fraction enriched in the homolog with one extra methyl suggests that the methyl group is located on position 6 of the @(1+3)-linked gluose unit. It seems unlikely that this methyl group is introduced by the same enzyme that catalyzes methylation of the a( 1 4 ) -l i n k e d glucose units, and the possibility exists that this minor component is modified to serve a specific function in the cell. Because the MGP molecule is also modified by esterification with several short-chain acyl groups (acetyl, propionyl, isobutyryl, succinyl, and octanoyl), it is apparent that individual molecules could be tailored to serve a variety of specialized roles related to the general lipid-binding function. In our search for 6-0-methylglucose lipopolysaccharide precursors, we have assumed that they would be weakly acidic glucans because of the presence of the glyceric acid and that some of the glucose would be methylated at position 6 or 3, or at both positions. The search was complicated because mycobacteria contain other acidic and partially methylated glucans, such as the pyruvylated 3-0-methylglucose-containing pentaglucoside (26). On gel filtration, this material is eluted in the position of a saccharide with 15 hexoses, so it could easily confuse any study in which the characterization was based on the incorporation of radioactive methyl groups. From the FAB mass spectra, however, the ion related to this glycolipid (m/z 735) was easily recognized as such.
Although the evidence is limited, we believe it supports our conclusion that the presumed precursors are not degradation products of MGP. First, previous attempts to demonstrate the existence of enzymic activities in M. smegmatis cells that could degrade radiolabeled MGP gave negative results (27). Second, the pattern of oligosaccharide structures we have found is not that expected for the amylolytic degradation of MGP as observed with the Rhizopus amylase (14), because intermediates at m/z 1957 and 2133 should not be formed.
Since most of the fractions we collected from the Bio-Gel P-4 column encompassed a size range of several hexose units, we would expect multiple ions in the mass spectra and that the smaller ions of one fraction should overlap with the larger ions of a neighboring fraction, which was observed. Although FAB mass spectra can yield fragment ions of the parent compounds, the multiple ions we observe in each spectrum reflect the presence of real components. This conclusion is supported by the relative intensities of the various ions and