Structural Characterization of the O-antigenic Polysaccharide of the Lipopolysaccharide from Rhizobium etli Strain CE3 A

The O-antigenic polysaccharide of the Rhizobium etli CE3 lipopolysaccharide (LPS) was structurally characterized using chemical degradations (Smith degradation and b -elimination of uronosyl residues) in combination with alkylation analysis, electrospray, and matrix-assisted laser desorption ionization-time of flight mass spectrometry, tandem mass spectrometry, and 1 H COSY and TOCSY nuclear magnetic resonance spectroscopy analyses of the native polysaccharide and the derived oligosaccharides. The polysaccharide was found to be a unique, relatively low molecular weight glycan having a fairly discrete size, with surprisingly little variation in the number of repeating units (degree of polymerization 5 5). The polysaccharide is O -acetylated and contains a variety of O -methylated glycosyl residues, rendering the native glycan somewhat hydrophobic. The molecular mass of the major de- O -acetylated species, including the reducing end 3-deoxy- D - manno -2-octulosonic acid (Kdo) residue, is 3330 Da. The polysaccharide is comprised of a trisaccharide repeating unit having the structure 3 4)- a - D -Glc p Polyga- lacturonic or malto-oligosaccharides calibration standards. High Pressure Liquid Chromatography Fractionation of Oligosaccha-rides— Oligosaccharides from the O-chain by b separated as described and then fractionated using a Vydac 5- m m monomeric C18 semi-preparative reverse phase (10 a linear gradient (0–30%) of methanol in containing 0.05% trifluoroacetic acid with a flow rate of 3 ml/min. Elution was an evaporative light scattering detector (Richard 9:1 in 90% of eluant was diverted from the detector and collected into fractions for structural analysis. glycosyl residues glycosides

(23, where QuiNAc is N-acetylquinovosamine (2-Nacetamido-2,6-dideoxyglucose). The reducing end Kdo residue links the O-chain polysaccharide to the core region oligosaccharide, resulting in a unique location for a Kdo residue in LPS, removed four residues distally from the lipid A moiety. Structural heterogeneity in the O-chain arises mainly from the O-acetyl and O-methyl substitution. Methylation analysis using trideuteriomethyl iodide indicates that a portion of the 2,3,4-tri-Omethylfucosyl capping residues, typically 15%, are replaced with 2-O-methyl-and/or 2,3-di-O-methylfucosyl residues. In addition, approximately 25% of the 3,4linked branching fucosyl residues and 10% of the 3-linked fucosyl residues are 2-O-methylated. A majority of the glucuronosyl residues are methyl-esterified at C-6. These unique structural features may be significant in the infection process.
The Rhizobium are Gram-negative bacteria capable of forming a nitrogen-fixing symbiosis with legumes in a host-specific manner. The initial stages of this process have been fairly well studied and involve an exchange of signal molecules (e.g. flavonoids and nod factors) (1)(2)(3)(4) between the bacterium and plant, leading to the induction of specific bacterial and plant genes. Relatively little is known about the molecular events occurring during the later stages of infection, including the subsequent development of the infection threads and bacteroids. Morphological studies have shown that the rhizobial cell surface is in contact with the plant cell surface throughout these stages, suggesting a prominent role for cell surface macromolecules in these events.
Rhizobium cells cultured as free-living bacteria express a variety of glycoconjugates on the cell surface, including the lipopolysaccharides (LPS), 1 capsular polysaccharides (K-antigens), and the extracellular polysaccharides (5)(6)(7). The LPS are major structural components of the rhizobial outer membrane and the dominant antigens of the rhizobial cell surface; antibodies raised to whole rhizobial cells are directed most strongly to the LPS (5, 8 -10). Studies with monoclonal antibodies directed to the O-chain moiety of various rhizobia have revealed that different types of LPS can be expressed by a single Rhizobium strain and that this expression depends on the in planta or physiological environment (11)(12)(13)(14)(15)(16). The majority of these antigenic changes appear to occur in the O-chain portion of the LPS and are associated with the different stages of symbiotic infection (11, 12, 14 -18). In addition, recent studies have shown that certain LPS structural changes can be induced by plant flavonoids, indicating that LPS expression can be modulated by nod/nol gene products (19 -21). In one example, an Rhizobium etli LPS O-chain epitope was suppressed by adding bean root or seed exudate to the growth medium (22), and the active compound was identified as an anthocyanin (21). Other studies have examined rhizobial LPS mutants and their symbiotic phenotypes and have shown that mutants that lack the O-antigen portion of their LPS or that contain truncated O-chains are unable to form normal infection threads (10,23,24) and/or are unable to invade the root nodule cells (25)(26)(27).
These studies have led to the general hypothesis that complete LPS molecules, containing a polymeric O-chain and intact corelipid A, are required for normal symbiosis to occur (24,28,29).
The structure of the R. etli-Rhizobium leguminosarum common core region has been elucidated (30 -32) and was found to have an entirely different structure from the typical core regions of enterobacterial LPS, including distinct linkage and branching patterns, the absence of both heptose and phosphate, an abundance of galacturonic acid, and notably, the location of a Kdo residue in the O-chain attachment region, removed four residues distally from the lipid A moiety. The lipid A moieties of these rhizobial LPS also differ significantly from the enterobacterial lipid As (33). The R. etli and R. leguminosarum lipid As are devoid of phosphate, have trisaccharide backbones containing uronic acid, glucosamine, and 2-aminogluconic acid, and contain the long chain fatty acid, 27hydroxyoctacosanoic acid, not found in the enteric lipid As. Like the enterobacterial LPS, the core-lipid A regions of the rhizobial LPS tend to be conserved within each Rhizobium genus, whereas structural diversity is typically found in the O-specific polysaccharide moiety (6,8,29). In the case of R. etli and R. leguminosarum, the structurally variable O-chains are strain-specific and are the dominant antigenic determinants (8 -10). In the majority of cases the structures of these epitopes have not been defined, and it is difficult to draw precise conclusions about their alterations or the significance of these alterations with regard to nodulation. In this report we describe the structure of the O-antigenic polysaccharide portion of the R. etli CE3 LPS. With the structure of this O-chain, the complete glycosyl sequence of the R. etli LPS, including the core region (30) and lipid A (33), is now known.

EXPERIMENTAL PROCEDURES
Growth of Bacteria-R. etli CE3, the streptomycin-resistant derivative of wild-type strain CFN42, was grown in tryptone/yeast extract supplemented with Ca 2ϩ as described previously (10). Bacteria were harvested by centrifugation at late log/early stationary phase.
Isolation of Lipopolysaccharide and the O-specific Polysaccharide-Lipopolysaccharides were extracted using the hot phenol/water procedure (10) and were analyzed by polyacrylamide gel electrophoresis (PAGE) in the presence of deoxycholate (34,35). Water layers containing LPS were dialyzed and treated sequentially with ribonuclease, deoxyribonuclease, and proteinase K and then redialyzed and subjected to size exclusion chromatography (SEC) in triethylamine/acetic acid buffer (10). The O-polysaccharide was isolated from the SEC-purified total LPS (both smooth and rough forms) by mild acid hydrolysis in 1.0% acetic acid at 105°C for 2 h. The hydrolysates were subjected to freeze/thaw (three repetitions), and the precipitated lipid A was removed by ultracentrifugation at 160,000 ϫ g for 1 h. The supernatants, containing the solubilized O-polysaccharide and the core region oligosaccharides, were passed through a column of polymyxin-agarose to remove remaining traces of free lipid A and intact LPS (30). The flow-through eluants were then chromatographed on Bio-Gel P-2 (45-90-m fine, 1.5 ϫ 90 cm) equilibrated in 50 mM ammonium formate, pH 6.5, to separate the O-chain from the core oligosaccharides.
Analysis of Glycosyl Residues-Carbohydrate compositions of the O-polysaccharide and derived fractions were determined by preparation of the TMS methyl glycosides with GLC-MS (electron impact) analysis (36) using a 30-m DB-5 fused silica capillary column (J & W Scientific). Carbohydrate identities and the location of endogenous O-methyl groups were also determined by analysis of the alditol acetate derivatives, using a 30-m SP-2330 capillary column (Supelco). Alditol acetates were prepared by hydrolysis in 2 M trifluoroacetic acid (121°C, 2 h), followed by reduction (NaBD 4 ) and acetylation (acetic anhydride/trifluoroacetic acid) (37). Acidic glycosyl residues were analyzed as the alditol acetates by converting the carboxyl groups to methyl esters (1 M methanolic HCl, 80°C for 2 h), followed by carboxyl reduction with NaBD 4 in water, trifluoroacetic acid hydrolysis, and conversion to the alditol acetates (33). Authentic N-acetylquinovosamine (QuiNAc) was prepared by chemical synthesis by Dr. R. Hollingsworth and provided as a gift. An authentic standard of 6-deoxy-L-talose was obtained from a streptococcal cell wall glycan (38).
Linkage analysis of neutral sugars was performed by permethylation (Hakomori method), conversion to the partially methylated alditol acetates (PMAAs) (36), and GLC-MS analysis. The Kdo and uronic acid linkages were identified by sequential permethylation, reduction of the carboxymethyl groups with lithium triethylborodeuteride (Aldrich), mild hydrolysis (0.1 M trifluoroacetic acid, 100°C, 30 min) to cleave ketosidic linkages, reduction (NaBD 4 ), normal hydrolysis (2 M trifluoroacetic acid, 121°C, 2 h), and conversion to the PMAAs (30). Methylations using trideuteriomethyl iodide were also performed to confirm sugar identities and the locations of endogenous O-Me ether groups. ␤-Elimination in conjunction with linkage analysis was performed by treating permethylated samples with 2 M dimethyl sulfoxide anion (potassium salt) in Me 2 SO for 8 h; the elimination products were then ethylated using ethyl iodide, and the products were converted to the corresponding partially methylated, ethylated alditol acetates and analyzed by GLC-MS (33). Where necessary, the identities of TMS methyl glycoside and PMAA derivatives were confirmed by chemical ionization-MS, using a 30-m DB-1 column and ammonia as the reactant gas. Chemical Modifications and Smith Degradation-De-O-acetylation of the O-specific polysaccharide was performed in 10 mM NaOH at 4°C for 10 h. The progress of the reaction was followed by 1 H NMR and MALDI-TOF mass spectrometry. Total ␤-elimination (depolymerization) was performed on portions of the native or de-O-acetylated polysaccharide using the following procedure. The polysaccharide (23 mg) was reduced with NaBD 4 (10 mg/3 ml 0.3 M NH 4 OH) overnight to stabilize the reducing end Kdo residue. The polysaccharide was then methyl-esterified by treating 3 times with methanol containing 1 to 2 drops of aqueous 1 M HCl and evaporating to dryness. The fully esterified polysaccharide was then treated with aqueous NaOH (0.25 M, 38°C, 18 h), and the products were chromatographed by (SEC) using Bio-Gel P-10 (1.5 ϫ 92 cm) in water. Further fractionation was performed on Bio-Gel P-4 (Ͻ45 mm, 1 ϫ 120 cm) or P-2 (Ͻ40 mm, 1 ϫ 120 cm). Column eluants were monitored for neutral carbohydrate using the phenol sulfuric acid assay (36) and for the unsaturated ⌬ 4,5 -hexuronosyl residues by measuring absorbance at 232 nm. Appropriate fractions containing oligosaccharides were combined and analyzed by ESI-MS, MALDI-TOF MS, and NMR as described below.
The Smith degradation (39) was performed by dissolving a sample of the de-O-acetylated O-chain polysaccharide in 50 mM sodium acetate buffer, pH 4.5, mixing with an equal volume of 0.2 M NaIO 4 , and maintaining at 4°C for 72 h in the dark. Excess periodate was destroyed (ethylene glycol), and the sample was reduced with NaBH 4 with the solution adjusted to pH 7.5. The resulting oxidized-reduced glycan ("polyol") was purified by chromatography on Bio-Gel P-10 and lyophilized. The polyol was subjected to mild hydrolysis (0.5 M trifluoroacetic acid, 23°C, 48 h); the acid was removed by evaporation at room temperature, and the products were separated on Bio-Gel P-2. The resulting oligosaccharides were analyzed for structural information.
Mass Spectrometry-Electrospray ionization mass spectrometry (ESI-MS) was performed on a SCIEX API-III triple-quadrupole mass analyzer (PE/SCIEX Thornhill, Ontario, Canada) operated in the positive ion mode with an orifice potential of 35-50 V (15). Spectra are the accumulation of 10 -60 scans collected over the m/z range 200 -2400 with a mass step of 0.2-1.0 atomic mass units at 1 ms/step. All watersoluble samples were desodiated by dialysis, followed by treatment with a strong cation exchange resin (H ϩ form), and then filtered using 0.45-m nylon spin filters and stored in polypropylene tubes. Underivatized samples were analyzed at a concentration of 2 g/l with a flow rate of 3-5 l/min using a solution of 15% v/v methanol in deionized water containing 0.5% v/v acetic acid. Permethylated polysaccharide samples were analyzed in 100% methanol containing 1% ammonium formate. Tandem mass spectrometry (MS/MS) was performed on the SCIEX instrument by selecting a parent ion for collision-induced dissociation using argon as collision gas. Matrix-assisted laser desorption ionization (MALDI)-mass spectrometry was performed on an LDI 1700XP time-of-flight (TOF) spectrometer (Linear Scientific, Reno, NV) in the positive and negative modes, using a matrix of 100 mM 2,5-dihydroxybenzoic acid in 90% methanol. The instrument was operated at an accelerating voltage of 30 kV and an extractor voltage of 9 kV. The sample was ionized with a nitrogen laser ( ϭ 337 nm) with a pulse width of 3 ns and 4 -7.5-J pulse. The following settings were used: detector sensitivity, 1275 mV full scale; voltage, 4.75 kV. Mass spectra were recorded over a m/z range of 1-50,000; spectra are the summation of 50 -250 acquisitions. Polygalacturonic acid or malto-oligosaccharides were used as calibration standards.
High Pressure Liquid Chromatography Fractionation of Oligosaccharides-Oligosaccharides derived from the O-chain by ␤-elimination were separated by SEC as described above and then fractionated using a Vydac 238TP510 5-m monomeric C18 semi-preparative reverse phase column (10 ϫ 250 mm) using a linear gradient (0 -30%) of methanol in water, containing 0.05% trifluoroacetic acid with a flow rate of 3 ml/min. Elution was monitored using an evaporative light scattering detector (Richard Scientific), using a 9:1 split in which 90% of the eluant was diverted from the detector and collected into fractions for structural analysis.
NMR Spectroscopy-1 H spectra were recorded with a Varian 300 or 600 spectrometer in D 2 O at 35°C using acetone as internal standard, ␦ 2.22. Two-dimensional NMR spectra (COSY and TOCSY) were performed using the Varian software.
Glycosyl Configuration-The enantiomeric configuration of the glycosyl residues was determined by GLC-MS analysis of the TMS (Ϫ)-2butyl glycosides of the polysaccharide in comparison with the authentic sugars (40). The Kdo was assumed to be in the D-configuration.

RESULTS
Isolation, Molecular Size, and Composition Analysis-Phenol/water extraction of R. etli strain CE3 typically allows recovery of both the smooth and rough LPSs in the water layer (ϳ800 mg of LPS/50 g of lyophilized cells), as shown in the deoxycholate-PAGE profile in Fig. 1, inset. MALDI-TOF-MS analysis of these LPSs is also shown in Fig. 1. It can be seen that the major MALDI component, corresponding to the LPS-I band, has an apparent molecular mass centered at around 6600 atomic mass units, using neutral maltooligosaccharides as the TOF calibrant. The molecular mass of this smooth LPS is consistent with the previously characterized masses for the R. etli lipid A moiety (major species 2 1987 atomic mass units (33)), and the core region oligosaccharide (major species 3 1531 atomic mass units) (30). These results suggested an approximate size for the O-chain of around 3100 atomic mass units.
The LPS was purified and subjected to mild acid hydrolysis, which selectively cleaves the Kdo ketosidic linkages yielding the O-chain polysaccharide, the core oligosaccharides, and the free lipid A. The released O-chain and core region oligosaccharides each carry a Kdo residue at the reducing end that can undergo a variety of degradative reactions as discussed below.
Following hydrolysis and lipid A removal, the soluble products were separated on Bio-Gel P-2, yielding the O-specific polysaccharide fraction (PS) at the void volume, and the lower molecular weight core region oligosaccharide fraction (OS) (data not shown). As shown in Fig. 2A, rechromatography of the PS fraction on Bio-Gel P-10 yielded a single symmetrical peak having a K av ϭ 0.18. MALDI-TOF-MS analysis of this PS revealed three major components, with an approximate molecular mass around 3000 atomic mass units (Fig. 2B). Inspection of these components showed each to be broad, with small in-  (41,42) The glycosyl composition of the PS fraction is reported in Table I. Analysis of the TMS methyl glycosides revealed that glucuronic acid, fucose, and an unidentified 3-O-methyl-6-deoxyhexose were the major components, along with lesser amounts of mannose, N-acetylquinovosamine, Kdo, and several partially O-methylated sugars. The PS fraction did not contain any fatty acids, N-acetylglucosamine, galacturonic acid, or other previously identified lipid A or core oligosaccharide components (30, 33), indicating that the PS consisted entirely of the O-chain polysaccharide. Analysis of the alditol acetates indicated that the 3-O-methyl sugar had a retention time identical to that of 3-O-methyl-6-deoxytalose. In an earlier study, it was reported that the total LPS from these cells contained 3-Omethylrhamnose, presumably the O-chain component (43). However, in the course of our present study it became clear that this component has a slightly different retention time from that of 3-O-methylrhamnose, when derivatized by either the TMS methyl glycoside, alditol acetate, or partially methylated alditol acetate procedures. The derivatives from this R. etli O-chain component had retention times identical to those from an authentic standard of 6-deoxytalose (38) but differed from the corresponding derivatives prepared from rhamnose, fucose, or quinovose. The O-polysaccharide contained a small amount of a dideoxy amino sugar, a 2-N-acetamido-2,6-dideoxyhexose, identified by the electron impact and chemical ionization mass FIG. 1. Positive ion MALDI-TOF mass spectrum and deoxycholate-PAGE analysis (inset) of the total R. etli LPS, isolated from the water layer of phenol/water extracts by Sepharose 4B chromatography. The strain CE3 LPSs consists of distinct smooth and rough forms (10,29). The LPS-I species contains the intact O-chain; LPS II is a rough LPS that lacks N-acetylquinovosamine and the O-chain repeating unit residues. LPS III and IV appear to carry truncated O-chains (one repeating unit) based on immunoblot results (10,29). In the mass spectrum, the apparent size heterogeneity of these LPS is due, in part, to fatty acid heterogeneity in the lipid A, to glycosyl heterogeneity in both the core region and the O-chain, and presumably to sodium adduct formation. spectra of its corresponding TMS (m/z 173, 204, 217, 247, 259, 316 inter alia), alditol acetate (m/z 85, 103, 145, 159, 201, 231, 242, 260, 302), and PMAA derivatives (shown below). This residue was determined to have the gluco configuration, since the retention times of these derivatives were identical to those obtained from authentic N-acetylquinovosamine. In addition, the 1 H ring proton connectivities observed during TOCSY analysis of oligosaccharides derived from the O-chain support the assignment of the gluco configuration for this sugar, as discussed below. GLC-MS analysis of the TMS (Ϫ)-2-butyl glycosides derived from the polysaccharide indicated that all fucose residues were of the L-configuration, and the glucuronic acid and mannose residues were of the D-configuration. The configurations for N-acetylquinovosamine, 3-O-Me-6-deoxytalose, 2,3,4-tri-O-Me-fucose, and the other endogenously methylated glycosyl residues could not be clearly assigned. Another interesting feature of the O-chain is that although it contains substantial glucuronic acid (Table I), most of the native polysaccharide (ϳ70%) fails to bind to DEAE-Sephadex, and the remaining 30% binds only weakly, due to the single Kdo resi-due that survives the mild acid hydrolysis (data not shown). This indicates that the GlcA carboxyl groups are blocked by some moiety.
A portion of the O-chain was subjected to de-O-acylation to allow accurate molecular mass assignment and facilitate chemical degradations, e.g. periodate oxidation. It was found, however, that the typical conditions used to effect de-O-acylation (e.g. 0.1 M NaOH, 1 h at room temperature) resulted in extensive depolymerization of the polysaccharide, due to base-catalyzed elimination. This was confirmed by the appearance of a  a Endogenously methylated residues were identified by comparison of their mass spectra and retention times with authentic standards generated by partial methylation.
b Ratios are normalized to mannose (Man) and are determined from the GLC-MS total ion current peak areas with response factor correction from authentic standards. 2,3,4-MeFuc, 2,3,4-tri-O-methylfucose; UV maximum at 232 nm (not shown) and subsequently by MS and NMR analyses discussed below. It thus became necessary to find suitable conditions for complete O-acetyl removal while allowing the recovery of an intact polysaccharide. Graded treatments with alkali under a variety of conditions were tried on a small scale, and the reactions were monitored by 1 H NMR (Fig.  3B) and MALDI-TOF MS (Fig. 4). It was found that 10 mM NaOH at 4°C for 10 h was sufficient to result in complete de-O-acetylation without depolymerization. It can be seen that this resulted in a slight overall reduction in molecular mass accompanied by a sharpening of the molecular ions, due to the loss of O-acetyl heterogeneity (Figs. 4 and 2A). Removal of the O-acetyl groups also allowed accurate molecular mass analysis by ESI-MS as shown in Fig. 5A. In the ESI-MS spectrum, the three major molecular species previously identified by MALDI-TOF are clearly resolved and are shown to differ by 202 and 187 atomic mass units, corresponding to the incremental masses of anhydro Kdo and QuiNAc residues, respectively. Hence, the O-polysaccharide consists of three major species that differ by loss of a Kdo and a QuiNAc residue. One possible explanation for this is a sequential loss of these labile residues from the reducing end of the molecule, during either the mild acid treatment or the de-O-acylation reaction, conclusions that were further substantiated by experiments described below. 3) The results support the likelihood that a QuiNAc residue is penultimate to the reducing end Kdo, as previously reported for the R. etli mutant CE359 LPS V (30). These results also argue against the possibility that strain CE3 synthesizes any O-chains that lack a QuiNAc residue, since a molecular species lacking only a QuiNAc residue was not observed during ESI-MS analysis.
In contrast, methyl group heterogeneity was observed at the fucosyl residues. Examination of the trimethyl derivative arising from the fucosyl residue(s) indicated that approximately   dues present in the glycan and listed in Table I arise from the branch point fucosyl residues.
Glycosyl Sequence Determination of the Repeating Oligosaccharide-The arrangement of substituents at the branching 3,4-linked fucosyl residue was investigated by ␤-elimination of the uronosyl residues in combination with methylation and ethylation analysis. The polysaccharide was methylated and then treated with the potassium dimethylsulfinyl anion. The ␤-elimination products were ethylated and then hydrolyzed and converted to the corresponding partially methylated, ethylated alditol acetates (Table II)  Other alternatives are possible for the tetrasaccharide unit. 3-Linked Man was also detected in the total mixture of ␤-elimination products (Table II); however, glycosyl composition and linkage analysis of the native O-chain indicated that this res-idue occurs to only a small extent and therefore is probably not part of a repeating structure. Rather, the 3-Man is part of a non-repeating sequence that occurs at the reducing end of the polysaccharide (described below).
To distinguish between the structural alternatives, the Smith degradation was applied to the de-O-acetylated glycan. A sample was treated with periodate and reduced with borohydride, and the reaction was chromatographed on Bio-Gel P-10. The oxidized-reduced product (polyol) eluted as a broad, tailing peak having a K av ϭ 0.47 (not shown). A sample of the polyol was isolated and subjected to composition and linkage analysis (Tables I and II, respectively). As expected, only the 4-linked glucuronosyl residues were susceptible to the periodate oxidation. A substantial portion of the Kdo also survived the periodate treatment, suggesting that this terminal Kdo residue was present as a 2,7-anhydro Kdof, rather than a 2,5-anhydro pyranose, which would be susceptible to oxidation. The remainder of the polyol was subjected to mild hydrolysis, constituting the Smith degradation, and the oligosaccharide products were isolated on Bio-Gel P-4. The major oligosaccharide fraction was analyzed by ESI-MS/MS (Fig. 6A) which revealed a major oligosaccharide (parent ion m/z 457.5; calculated M ϩ H ϩ 457.4) consistent with Structure 3.
A second Smith derived oligosaccharide (M ϩ H ϩ ϭ 471.5), which co-eluted with the major oligosaccharide on P-4, was identified by ESI-MS/MS as containing two contiguous O-methyl 6-deoxyhexosyl residues linked directly to a terminal tetronic acid aglycon (Fig. 6B). These results are consistent with an oligosaccharide having the following Structure 4.
This glycosyl sequence is also predicted from the linkage analysis using trideuteriomethyl iodide. Both of these oligosaccharides are consistent with a trisaccharide repeating unit, as shown in Structure 1 above. The occurrence of a methyl ester in each of these oligosaccharides presumably indicates the C-6 substituent of the glucuronosyl residues. As discussed previously, methylation analysis of the native polysaccharide revealed substantial 3-linked Fuc (Table II), suggesting the possibility of a tetrasaccharide repeating unit. However, as shown below, the intact polysaccharide contains only one 3-linked Fuc residue, which originates from the non-repeating glycosyl sequence at the reducing end of the molecule. The additional 3-linked Fuc reported in Table II appears to arise from ␤-elimination and subsequent degradation of some of the esterified GlcA residues during the methylation with dimsyl anion. This also results in a slight decrease in the amount of 3,4-linked Fuc, with respect to the terminal 6-deoxy-Tal and 4-linked GlcA derivatives (Table II). Clearly, the results of methylation analysis of intact polysaccharides can only approximate the structures deduced by other means, such as MS/MS analysis of the Smith-derived fragments.
Oligosaccharides derived from the non-repeating sequences at both the reducing and nonreducing ends of the O-chain were also detected during the Smith procedure; however, their yields were relatively low, presumably because they are not repeated sequences and because some were lost during the work up of the polyol. Several repetitions of the oxidation-reduction procedure, including chromatography steps, were required to preclude inter-residue hemiacetal formation and thus ensure complete oxidation. A more direct procedure for the isolation of the reducing and nonreducing end oligosaccharides was employed using ␤-elimination, as described below.
Glycosyl Sequence of the Nonreducing and Reducing Ends-Portions of the O-chain polysaccharide were pre-reduced with borodeuteride and subjected to base-catalyzed ␤-elimination as described under "Experimental Procedures." The oligosaccharide products were chromatographed on Bio-Gel P-10 (Fig. 7). The P-10 profile shows clearly that these depolymerization products are of lower molecular weight than that of either the native or de-O-acetylated polysaccharides (compare Fig. 2A) and have absorbance at 232 nm, characteristic of a 4-O-substituted hexuronate ␤-elimination product, i.e. 4-deoxyhex-4enopyranosiduronate residues. The major fraction, P-10Ϫ2 (K av ϭ 0.59), was isolated and chromatographed on Bio-Gel P-4; however, the material failed to undergo further separation and yielded a poorly resolved, peak/shoulder that migrated near the void volume of the P-4 column (not shown). Samples were taken from both the leading and tailing edges and were analyzed for glycosyl composition (Table I) and by ESI-MS. Five molecular ions were observed, indicating that this fraction consisted of a mixture of five oligosaccharides, designated OS-1 to OS-5. The major molecular species and their proposed for-mulas are summarized in Table III. The relative proportion of these molecular ions varied between the leading and tailing edge of the P-4 peak: the leading edge (higher molecular weight) was enriched in OS-2-5, and the tailing edge (lower molecular weight) was enriched in OS-1. As will be shown below, OS-1 originates from the nonreducing end of the Ochain, and OS-2-5 originate from the reducing end of the molecule. The ESI-MS spectrum showing the molecular ions for OS-1-5 is shown in Fig. 8.
To establish the glycosyl sequence of these oligosaccharides, the major molecular ions were selected for CID-MS/MS. For most of the spectra, the fragment ions consisted of Y-type ions resulting from glycosidic cleavage with charge retention on the reducing end, and B-type ions, arising from oxonium ion formation with charge retention on the nonreducing end (44). Second generation fragment ions resulting from double cleavages were also observed in several of the spectra. The ESI-MS/MS spectrum of OS-1 is shown in Fig. 9. The results show that OS-1 contains the 2,3,4-tri-Me-Fuc residue, and OS-1 thus represents the nonreducing end of the glycan. Glycosyl composition and linkage results previously indicated that this residue occurs only once in the glycan, as a terminally linked or "capping" residue. One of the reasons for the low recovery of this oligosaccharide during the Smith degradation is that most of the glucuronic acid was oxidized by periodate, leaving only the single 2,3,4-tri-Me-fucosyl residue as a representative of this nonreducing end "sequence." During the ␤-elimination, a portion of the glucuronosyl residues survived the base treatment, allowing detection of this intact sequence (OS-1).
Examination of oligosaccharides OS-2-5 (Figs. 8 and 10 and Table III) revealed them to be a series of related structures that differ in the nature of their reducing and nonreducing end residues. OS-4 and OS-5 terminate at the nonreducing end with a 4-deoxyhex-4-enopyranosiduronate residue, resulting from the ␤-elimination. Since the glycosidic linkage of these unsaturated residues is labile, a portion of these residues is lost during the ␤-elimination and isolation, yielding OS-2 and OS-3. Hence, OS-4 and OS-5 are each 158 mass units greater than OS-2 and OS-3, respectively, corresponding to the interval FIG. 7. Bio-Gel P-10 fractionation of the ␤-elimination products derived from the O-chain. Oligosaccharides derived from both the nonreducing and reducing ends of the O-chain co-eluted in P-10Ϫ2 (K av ϭ 0.59, compare Fig. 2). The oligosaccharides in this fraction, designated OS-1 to OS-5, were further separated by Bio-Gel P-4 chromatography and reverse phase HPLC and analyzed for glycosyl composition and linkage by GLC-MS, ESI-MS/MS, and 1 H NMR. Fraction P10-1 contained partially degraded O-chain fragments, characterized by some intact repeating unit sequences, in which some of the glucuronosyl residues did not undergo ␤-elimination. Fraction P10-3 consisted of 3-O-Me-6-deoxytalose, fucose, and unidentified degradation products arising from these residues due to ␤-elimination of the interior glucuronosyl residues. mass of the ⌬ 4,5 -hexuronate residue. In addition, OS-3 and OS-5 terminate at the reducing end with a Kdo alditol, resulting from borodeuteride reduction, and OS-2 and OS-4 terminate with a 2,7-anhydro Kdo residue, which is resistant to reduction. The mass difference between OS-2 and OS-3 is 21 atomic mass units, which is due to the replacement of an anhydro Kdo residue with a reducing Kdop (ϩ18 atomic mass units) reduced at C-2 with borodeuteride (i.e. CϭO 3 D-C-OH ϭ ϩ3 atomic mass units). An identical difference exists for the OS-4/OS-5 pair. Furthermore, prereduction of the polysaccharide with borohydride, rather than borodeuteride, resulted in a difference of 20 atomic mass units for these molecular species (i.e. OS-2 and OS-4 remained at m/z 1022.5 and 1180.5, respectively, whereas OS-3 and OS-5 were shifted to m/z 1042.5 and 1200.5, respectively) (data not shown). Further confirmation of the charge state and type of cationization (i.e. protonation) was obtained by the controlled addition of sodium, which resulted in an increase of 22 m/z units for each of these species. These results support the conclusions drawn from linkage and ESI-MS analysis of the intact polysaccharide, namely that both types of Kdo residues (anhydro and reducing) are present at the reducing end of the glycan. The MS/MS spectrum of OS-2 is shown in Fig. 10. Together with composition and linkage analyses, these results indicate that the reducing end of the polysaccharide contains the non-repeating sequence 33)-Fucp-(133)-Manp-(133)-QuiNAcp-(134)-Kdo-(23, where C-3 of the fucosyl residue is substituted by the first repeating unit in the polysaccharide. The oligosaccharide mixture (OS-1-OS-5) was fractionated by reverse phase HPLC, yielding a major fraction that contained OS-2 and OS-4 and a second fraction containing OS-1, OS-3, and OS-5, as determined by ESI-MS (data not shown). Further separation was not attempted, and the structurally related OS-2/OS-4 mixture was analyzed by 1 H COSY and TOCSY experiments (Fig. 11). The distinctive signal at ␦ 5.   ture values (45) suggests an ␣-linkage for this unsaturated residue, indicating that the glucuronosyl residues have the ␣-anomeric configuration in the intact polysaccharide. Five other ring systems were also identified (Fig. 11), and the chemical shifts of the anomeric protons for all of these glycosyl residues also suggests they are ␣-linked, with the exception of two resonances, at ␦ 4.55 and ␦ 4.62. The signal at ␦ 4.55 is attributed to a ␤-mannosyl residue, due to weak H-1/H-2/H-3 coupling, suggesting the manno configuration for this system. The signal at ␦ 4.63 is attributed to a ␤-N-acetylquinovosamine residue, since cross-peaks can be detected for C-6 methyl protons as well as all ring protons in this system. The chemical shifts for the QuiNAc ring protons are H-1 (␦ 4.62), H-2 (␦ 3.72), H-3 (␦ 3.36), H-4 (␦ 3.75), H-5 (␦ 3.52), and H-6 (␦ 1.31). Thus, the QuiNAc residue (6-deoxy-gluco configuration) shows coupling throughout all ring protons, whereas the other three 6-deoxyhexose methyl protons show strong coupling only from H-6 to H-5, characteristic of fuco and talo configurations.
The anomeric configuration of the Kdo residue was investigated indirectly, by examining the Kdo residues in a rough LPS synthesized by the R. etli mutant CE358. The LPS from this mutant contains a core oligosaccharide that terminates at the nonreducing end with the distal Kdo residue, which normally serves as the site of O-chain attachment in those LPS that contain an O-chain (30). The core region of this mutant, as with the core region of the parent strain, contains a total of three Kdo residues, including two in the interior portion of the core. 1 H COSY analysis of this de-O-acylated rough LPS showed intense signals for all Kdo methylene protons at ␦ 2.02 (H3a) and ␦ 1.78 (H3e), and the chemical shift difference (⌬ ϭ 0.24 ppm) is indicative of ␣-linked Kdop residues (31,46). DISCUSSION The glycosyl sequence of the O-antigenic polysaccharide from the R. etli CE3 LPS is shown in Fig. 12. The polysaccharide is characterized by a number of rather novel structural features for an O-chain, including a relatively low molecular weight (3330 Da), uniform size (five repeating units with little variation in the number of repeats), numerous endogenously Omethylated components (3-O-methyl-6-deoxytalose, 2,3-di-Omethylfucose, 2-O-methylfucose, and 2,3,4-tri-O-methylfucose), and O-acetylation. The latter features appear to confer a fairly high level of hydrophobicity to the glycan, a property that may be important during infection thread development and the establishment of symbiosis 4 (6,47). With the elucidation of this structure, the complete glycosyl sequence of the R. etli CE3 LPS, including the unique core region (30) and lipid A moiety (33), is now known for the laboratory-cultured bacterium. The results presented here also show that 3-O-methylrhamnose is not a glycosyl component of the O-chain repeating unit (rather, what was thought to be 3-O-methylrhamnose is actually 3-Omethyl-6-deoxytalose), as was suggested in earlier reports (6,43) that were based on incomplete preliminary data.
Studies with rhizobial LPS mutants and their symbiotic phenotypes have helped define our current understanding of the role of these molecules in symbiotic infection. These studies have led to the general observation that complete LPS molecules, synthesized in normal amounts and containing a polymeric O-chain, core oligosaccharide, and lipid A moiety, are required for effective symbiosis (24,28). Of recent interest from a structural and biosynthetic standpoint is the R. etli mutant CE166, (Ndv ϩ , Fix Ϫ ) which produces reduced amounts of a smooth LPS (LPS-I) containing a structurally modified O-chain. The symbiotic defect of this mutant can be partially suppressed by the incorporation of a multicopy plasmid carrying R. etli CE3 DNA, a modification that also appears to restore LPS synthesis to levels approaching that of the parent strain (23). Interestingly, both the mutant LPS and the suppressed strain LPS lack N-acetylquinovosamine, 5 indicating that the O-chain of these LPS is attached to the core region by a different glycosyl sequence than that reported here for the parent strain. The structure of this O-chain and the nature of its attachment to the outer core region are currently under study.
Studies with monoclonal antibodies directed to the O-chain moiety of various rhizobia have revealed that antigenic changes occur in these molecules during the different stages of symbiotic infection (11, 12, 14 -18) and in response to different physiological conditions established in culture (11)(12)(13)(14)(15)(16). In the majority of cases the nature of these structural changes has not been defined. Relatively modest changes, such as the extent or position of O-acetylation or O-methylation, could easily account for some of the alterations in antibody reactivity. Recently, Carlson et al. (47) surveyed some of the chemical changes occurring in R. leguminosarum LPS isolated from normally cultured cells and from bacteroids, and they compared these changes with cells cultured under different physiological conditions intended to mimic the nodule environment. 4 It was found that LPS obtained from nodule bacteria or those grown under physiological extremes were more hydrophobic than LPS extracted from normally cultured cells; the increased hydrophobicity was attributed to a higher proportion of longer chain fatty acids in the lipid A and a higher degree of O-acetylation or O-methylation of the O-chain sugars. Based on the partitioning of whole cells between organic and water layers, they proposed that the entire rhizobial cell surface became more hydrophobic 4 E. Kannenberg  The spectrum is characterized by consecutive Y-type fragment ions and a single B ion (B 5 ), all derived from the parent ion. During the electrospray MS/MS, much of the fragmentation appears to be directed by the amino sugar residue. Thus, the Y 1 ion (m/z 408), containing the amino sugar, is particularly stable. The B 5 ion, which terminates in the amino sugar, does not undergo further B-type fragmentation and instead gives rise to a series of second generation fragment ions (m/z 642, 496, 350, and 188) due to consecutive ␤-cleavages from the nonreducing end. when cells were isolated from nodules or grown under the extreme conditions. The exact nature of the structural changes occurring in these LPS (the major surface component) is currently under study.
The biosynthetic mechanism leading to the R. etli CE3 Ochain polysaccharide has not been examined; however, of the two major routes proposed for O-chain biosynthesis, (reviewed in Ref. 48), the "monomeric" mechanism, in which individual monosaccharide residues are transferred consecutively from the corresponding glycosyl donor (XDP-sugar) to the non-reducing end of the growing chain, appears to be the more likely. This mechanism has generally been associated with the uniformly sized O-chains characteristic of certain strains of Escherichia coli (49), Rhodospirillum rubrum (34), and others (48). This type of discrete size LPS/O-chain is frequently encountered in bacteria that synthesize a crystalline protein surface layer (S-layer), such as the eubacteria. Although the rhizobia are not known to possess an S-layer, the presence of O-chains of homogeneous size could allow the ordered assembly of other surface structures or in some way promote interaction with infection thread membrane components. Mechanisms regulating the chain length of O-chain polysaccharides have not been elucidated; however, the presence of the 2,3,4-tri-methylfucose residue at the nonreducing end of the R. etli O-chain would obviously preclude further elongation. Terminally linked Omethyl sugars have also been proposed as capping residues for other O-chain polysaccharides, specifically those believed to be synthesized by the monomeric mechanism, for example Klebsiella O5 (50), E. coli O8 (49), and Rhizobium loti NZP2213 (51).
The biological significance of the various partially O-methylated and O-acetylated residues is not immediately clear, aside from the obvious consideration of hydrophobicity. Many sugar methyl ethers, particularly the O-methyl ethers of 6-deoxyhexoses, are frequent O-chain components in soil bacteria including species of Rhizobium (43,51,52), Sinorhizobium (53), Bradyrhizobium (8), Agrobacterium (54), and others. The O-methyl ethers of rhamnose and fucose are most common; however, 2-O-methyl-6-deoxy-L-talose was identified as an O-chain component of R. loti strain NZP2213 (51) and in Burkholderia pseudomallei (55). The R. loti NZP2213 O-chain was found to be a homopolymer of non-methylated 1,3-linked 6-deoxy-Tal, in 30:1 ratio with the 2-O-Me-6-deoxy-Tal. The 2-O-Me-6-deoxy-Tal was proposed to be at the non-reducing end, essentially a capping residue. Non-methylated 6-deoxy-L-Tal was also recently characterized in the trisaccharide repeating unit in the R. leguminosarum bv. trifolii strain 24 O-chain, in which it was attached as a side chain ␣1,2-linked to rhamnosyl residues (56). The remaining residue of this O-chain repeating unit is the acidic sugar 3-deoxy-D-lyxo-heptulosaric acid (56). Interestingly, a Fix Ϫ mutant of this strain, 24AR, synthesizes a totally different O-chain that lacks 6-deoxy-Tal but is rich in heptose and O-methylheptose. 6 The only other well characterized Ochain repeating unit from a bean microsymbiont (R. tropici CIAT899) also contained 6-deoxy-Tal 3-linked to fucose (57). The 3-O-methyl ether of 6-deoxy-Tal was not previously reported as a component of any rhizobial LPS and has been encountered only infrequently, as the L-isomer in Stenotrophomonas maltophilia O-chains (58), and in the D-configuration in O-chains from certain photosynthetic bacteria (59).
Another interesting structural feature of the R. etli Opolysaccharide is the presence of the non-repeating sequence 33)-␣-L-Fucp-(133)-␤-D-Manp-(133)-␤-QuiNAcp-(134)-␣-Kdo-(23, which connects the repeating unit residues of the O-chain to the galactosyl residue of the core oligosaccharide. This Kdo residue is thus removed four residues distally from the lipid A moiety, resulting in a unique location for a Kdo residue in an LPS (30). The anomeric configuration of this distal Kdo residue, as well as that of the two Kdo residues in the inner core region, is ␣, suggesting that attachment of this residue falls within the domain of the typical core region biosynthetic machinery. Interestingly, the enzyme that transfers this Kdo residue and the gene that encodes this activity were recently described (60,61). Thus, the Kdo residues of both the enteric and rhizobial LPS core regions are ␣-linked, whereas ␤-linked Kdo is a frequent constituent of capsular or other cell surface polysaccharides (6,62), including the rhizobial Kantigens (6,63). It is also interesting to note that the rough LPS species LPS II, produced by the parent strain (see Fig. 1), lacks the Fuc, Man, and QuiNAc residues, although a portion of these R-LPS is indeed terminated with the distal ␣-Kdo residue. As a result of these observations and because of its contrasting chemical nature, the 3-linked QuiNAc residue was originally believed to be essential for O-chain elongation or attachment to the core region in the R. etli LPS. However, as discussed above, the R. etli mutant CE166 and its suppressed strain lack the QuiNAc residue yet are able to synthesize substantial amounts of a smooth LPS species, the O-chains of which appear to be different in structure from that reported here for the parent strain. In addition to the parent strain, the glycosyl sequence 3 Fuc3 Man3 QuiNAc3 Kdo3 was also detected, along with a single O-chain repeating unit, in the "semirough" LPS V component synthesized by the R. etli mutant CE359 (Ndv ϩ , Fix Ϫ ) (30). From a structural standpoint (i.e. the presence of hexose and QuiNAc) this sequence bears some resemblance to the outer core region modifications found in numerous non-enterobacterial rough LPS and lipooligosaccharides, the outer core regions that frequently contain hexose and amino sugars, including D-QuiNAc (64) and its isomers (65,66). However, the genomic locations of the glycosyltransferases responsible for this R. etli sequence and their relative proximity to known O-chain or core region encoding regions have not been examined.