INTRODUCTION

Lipopolysaccharide (LPS)¹ is a major component of the outer leaflet of the outer membranes of Gram-negative bacteria (1, 2, 3). It is composed of three covalently linked domains: lipid A, a hydrophobic glucosamine-based anchor; core oligosaccharide, a non-repeating structure consisting of inner and outer regions; and O-antigen, a distal, repeating oligosaccharide (1, 2, 3, 4, 5). Mutants lacking O-antigen and most core sugars are viable (3, 6, 7). Lipid A and the 3-deoxy-D-manno-octulosonic acid (Kdo) of the inner core are essential, and biosynthetic mutations must be isolated as conditional lethals (8, 9, 10, 11).

The E. coli inner core (Fig. 1) contains two Kdo residues attached to the 6-position of lipid A, and two L-glycero-D-manno-heptose residues (heptose) attached to the inner Kdo (1, 2, 3). Additional phosphate, phosphoethanolamine, and heptose (not shown) may be present in sub-stoichiometric amounts (1, 3, 5, 12). The inner core of the nitrogen-fixing bacterium Rhizobium leguminosarum contains at least two Kdo residues with linkages that are probably identical to those of E. coli (13, 14). Heptose is absent in the inner core of R. leguminosarum, but mannose is found attached to Kdo in an 1-5-linkage (Fig. 1) (14, 15). The R. leguminosarum core also contains one galactose and three galacturonic acid residues (Fig. 1) (14, 15). A plausible structure for the R. leguminosarum core is shown in Fig. 1. It is based on limited information (14, 15) and is less well characterized than that of Escherichia coli. The genes encoding the glycosyltransferases that assemble the core of R. leguminosarum have not been identified. Assays for enzymes of core biosynthesis in R. leguminosarum have not been reported.

Despite the divergent structures of the mature lipid A and core domains of E. coli and R. leguminosarum, both organisms employ the same seven enzymes to generate the LPS precursor Kdo₂-lipid IV_A (Fig. 2) (1, 4, 13, 16). We have therefore studied the further utilization of Kdo₂-lipid IV_A in extracts of R. leguminosarum to uncover unique aspects of the Rhizobium system (1, 17, 18). We now show that Kdo₂-lipid IV_A can serve as an acceptor of mannose residues derived from GDP-mannose in extracts of R. leguminosarum, but not E. coli. The mannosyl transferase is associated with the inner membrane. Synthetic ADP-mannose can substitute for GDP-mannose.

The incorporation of the first heptose (Fig. 1) into the core of E. coli, which resides at a position that may be occupied by mannose in R. leguminosarum, is accomplished by the action of the rfaC gene product, designated heptosyl transferase I (1, 3, 19, 20). RfaC transfers heptose from the proposed physiological donor ADP-heptose (20, 21, 22, 23) to position 5 of the inner Kdo of Kdo₂-lipid IV_A (Fig. 2, arrow) (1, 3). Despite the fact that the Salmonella typhimurium and E. coli rfaC genes have been cloned and overexpressed, no quantitative assay for RfaC has been reported (1, 3, 20).

As seen in Fig. 2, the structures of mannose and heptose are quite similar. All the chiral centers that they have in common are identical, but heptose contains one additional CH₂OH group. We now show that synthetic ADP-mannose can function as a substrate in place of ADP-heptose in the reaction catalyzed by E. coli or S. typhimurium RfaC. Although ADP-mannose has no known physiological function, it is commercially available, and its use should facilitate the characterization of heptosyl transferase I.

EXPERIMENTAL PROCEDURES

[-³²P]ATP and GDP-[U-¹⁴C]mannose were obtained from DuPont NEN. Silica Gel 60 (0.25 mm) thin layer chromatography plates were obtained from Merck. ADP-mannose and GDP-mannose were purchased from Sigma. All solvents were reagent grade. Radiochemical analysis of thin layer plates was performed with the Molecular Dynamics PhosphorImager model 425S and ImageQuant software. Plasmid isolation reagents (Wizard Minipreps) were obtained from Promega.

Wild type R. leguminosarum CE3 was obtained from D. Noel (Marquette University, Milwaukee, WI) (24). R. leguminosarum strains 24 (wild type parental) and 24AR (lacking the 4-phosphatase) (17) were obtained from R. Russa (Marie Curie Sklodowska University, Lublin, Poland) (25). Wild type R. meliloti 1021 was obtained from S. Long (Stanford University). E. coli K12 strain R477 has been previously described (26). E. coli K12 strains D21 (parental) and D21f2 (rfaC mutant) were obtained from the E. coli Genetic Stock Center (Yale University) (27, 28, 29). S. typhimurium LT2 strain SA3624 (harboring plasmid pKZ84/rfaC⁺) was obtained from the Salmonella Genetic Stock Center (University of Calgary, Alberta, Canada) (20).

Strains of Rhizobium were grown at 30 °C on TY medium (5 g of tryptone and 3 g of yeast extract per liter supplemented with 10 mM CaCl₂) with nalidixic acid at 20 µg/ml and streptomycin at 200 µg/ml (13, 17). E. coli and S. typhimurium were grown at 37 °C on LB medium (10 g of tryptone, 5 g of yeast extract, 10 g NaCl per liter) (30). With strains harboring the plasmid pKZ84, ampicillin was added at 50 µg/ml.

One liter of late log phase cells (A₆₀₀ = 1) were harvested in the cold (0-4 °C) by centrifugation at 6,000 × g for 15 min, washed once with ~500 ml of 50 mM HEPES, pH 7.5, centrifuged again, and resuspended in 10 ml of 50 mM HEPES, pH 7.5. The cells were broken by passage through a French pressure cell at 18,000 p.s.i., yielding a protein concentration of approximately 10 mg/ml. Cellular debris was removed by centrifugation at 6,000 × g for 15 min. Membranes were prepared by ultracentrifugation at 100,000 × g for 60 min. The membrane pellet was resuspended in ~1 ml of 50 mM HEPES, pH 7.5 (~10 mg/ml protein). The protein concentrations of the extracts, membranes, and cytosol were determined by the bicinchoninic acid assay (31) using bovine serum albumin as the standard.

[4-³²P]-Lipid IV_A was generated from [-³²P]ATP and disaccharide 1-phosphate precursor, using the E. coli 4-kinase in membranes of strain BR7 (32). Next, the labeled lipid IV_A was converted either to Kdo₂-[4-³²P]-lipid IV_A using purified E. coli Kdo transferase (16, 26) or to Kdo-[4-³²P]-lipid IV_A using a Hemophilus influenzae extract (33). The product was purified by preparative thin layer chromatography and stored at 20 °C as an aqueous dispersion (16, 26). Before each use, these substrates were subjected to ultrasonic irradiation in a water bath for 60 s.

Unless indicated, reaction mixtures (10-50 µl) contained 50 mM HEPES, pH 7.5, and 0.1% Triton X-100. When using ³²P-labeled lipid acceptor, the substrate concentrations were 2.5 µM (Kdo)₂-[4-³²P]-lipid IV_A (80,000 cpm/nmol) and 1 mM GDP-mannose (or ADP-mannose). When using the ¹⁴C-labeled substrate, the concentrations were 2 µM GDP-[U-¹⁴C]mannose (440,000 cpm/nmol) and 6.25 µM (Kdo)₂-lipid IV_A. Enzyme (0.05-1.0 mg/ml crude extract) was added to start the reaction, and the mixture was incubated at 30 °C for 1-60 min.

Reactions were stopped by spotting 5-µl portions of the reaction mixtures onto a Silica Gel 60 thin layer chromatography plate. After drying in a stream of cold air, plates were developed in the solvent chloroform, pyridine, 88% formic acid, water (30:70:16:10, v/v). The amount of product formed was calculated from the percent conversion of radioactive substrate (of known specific radioactivity) to product, which was determined after overnight exposure using a PhosphorImager.

R. leguminosarum membranes were separated by isopycnic sucrose gradient centrifugation following the procedure of Guy-Caffey et al. (34) for E. coli with the modification that the cells were not converted to spheroplasts prior to French pressure cell treatment. The protein content of each fraction was determined by the bicinchoninic acid assay (31) using bovine serum albumin as the standard, and turbidity was measured at A₆₀₀. The fractions were assayed for the inner membrane marker, NADH oxidase (35). The fractions were also assayed for the outer membrane marker, phospholipase A (36), using phosphatidylcholine as the substrate. Last, each fraction was assayed for mannosyl transferase activity using GDP-[U-¹⁴C]mannose as the donor (see above).

Plasmid DNA was isolated with the Promega Wizard miniprep kit. Transformation of plasmid DNA into competent cell lines was performed by high voltage electroporation using a Bio-Rad Gene Pulser II at 2.5 kV, 200 , and 25 microfarads.

RESULTS

Given that mannose is attached to the Kdo of the core of R. leguminosarum, but not of E. coli or R. meliloti, we assayed crude extracts from several strains for GDP-mannose-dependent glycosylation of Kdo₂-lipid IV_A. As shown in Fig. 3 (panel A, lanes 4, 6, and 8), extracts of several R. leguminosarum strains supported efficient conversion of Kdo₂-[4-³²P]-lipid IV_A to a more hydrophilic product, indicated as mannosyl-Kdo₂-[4-³²P]-lipid IV_A. Formation of this substance was absolutely dependent upon addition of GDP-mannose (Fig. 3, panel A, lanes 3, 5, and 7). Extracts E. coli and R. meliloti lacked the ability to form the GDP-mannose dependent product (Fig. 3, panel A, lanes 2 and 10), even upon prolonged incubation (not shown).

The putative product, mannosyl-Kdo₂-lipid IV_A, is derived from both Kdo₂-lipid IV_A and GDP-mannose, as evidenced by the generation of a ¹⁴C-labeled lipid product migrating in the same place when non-radioactive Kdo₂-lipid IV_A and GDP-[U-¹⁴C]mannose are employed as substrates (Fig. 3, panel B). As in panel A of Fig. 3, only those extracts made from R. leguminosarum cells contained enzymatic activity (Fig. 3, panel B, lanes 4, 6, and 8). Extracts of E. coli and R. meliloti (Fig. 3, panel B, lanes 2 and 10) were inactive. Omission of Kdo₂-lipid IV_A resulted in the complete absence of ¹⁴C-labeled product (Fig. 3, panel B, lanes 3, 5, and 7).

R. leguminosarum 24AR (25) was used in all subsequent enzymatic experiments because mannosyl transferase activity appears to be significantly higher in extracts of this strain (Fig. 3, panel B, lane 6). Additionally, 24AR lacks the 4-phosphatase activity (17) normally present in wild type R. leguminosarum extracts. Some degradation of GDP-mannose does occurs in R. leguminosarum extracts (material just above the GDP-[U-¹⁴C]mannose substrate in lanes 4-9 of panel B), but this reaction is independent of Kdo₂-lipid IV_A (lanes 5, 7, and 9).

The mannosyl transferase present in crude cell extracts is time and protein concentration dependent (Fig. 4), and it catalyzes quantitative mannosylation of 6.25 µM Kdo₂-[4-³²P]-lipid IV_A in the presence of excess (1 mM) GDP-mannose. The reaction has a pH optimum centered around 7.5. The activity is at least 10-fold lower at pH 6 (data not shown). The reaction is absolutely dependent upon the presence of a non-ionic detergent, like Triton X-100. The optimal concentration of Triton X-100 is 0.1-0.2%, provided the crude extract concentration is below 1 mg/ml. Both dithiothreitol and EDTA are slightly inhibitory. A preliminary kinetic characterization of the mannosyl transferase was performed. With crude extract as the enzyme source (Fig. 5), the apparent K_m for Kdo₂-lipid IV_A is 7.1 µM in the presence of excess GDP-mannose, and the apparent K_m for GDP-mannose is 4.3 µM in the presence of excess Kdo₂-lipid IV_A.

About 90% of the mannosyl transferase activity present in R. leguminosarum 24AR extracts is membrane associated, as indicated in Fig. 6, panel A. The specific activity of the membranes using the standard ¹⁴C assay is 850 pmol/min/mg, a 4-fold enrichment over extracts. Mannosyl transferase activity in cytosol is 16 pmol/min/mg. Membrane fractionation by sucrose density gradient centrifugation localizes the mannosyl transferase to the inner membrane (Fig. 6, panels B and C), as judged by NADH oxidase, which serves as a marker. Phospholipase A indicates the presence of outer membranes (panel B).

The specificity of the mannosyl transferase for Kdo₂-lipid IV_A as the acceptor was tested by comparing lipid IV_A, Kdo-lipid IV_A, and Kdo₂-lipid IV_A as substrates. As seen in Fig. 7, there was no detectable reaction with either lipid IV_A or Kdo-lipid IV_A under conditions that result in nearly quantitative mannosylation of Kdo₂-lipid IV_A. Apparently, both Kdo residues are required for substrate recognition, even though the mannose is thought to be attached to the inner Kdo. A similar requirement for the presence of both Kdo residues has been observed in the transfer of laurate to Kdo₂-lipid IV_A catalyzed by the product of the E. coli htrB gene (37, 38). However, given that the structure of the R. leguminosarum inner core is not fully established (14), it remains possible that the mannose is actually attached to the outer Kdo in R. leguminosarum. Further structural analyses will be required to resolve this issue.

Synthetic ADP-mannose has no known physiological function, but it is commercially available. As shown in Fig. 8 (lanes 3-5), extracts of the same strains of R. leguminosarum that glycosylate Kdo₂-[4-³²P]-lipid IV_A in a GDP-mannose dependent manner (Fig. 3) also function with ADP-mannose as the donor. This is not entirely surprising, since guanine and adenine containing nucleotides can often substitute for each other in metabolism. Extracts of R. meliloti are inactive with ADP-mannose (Fig. 8, lane 6), as with GDP-mannose.

ADP-mannose-dependent Glycosylation of Kdo₂-[4-³²P]-lipid IV_A in Extracts of rfaC⁺ E. coli

Extracts or membranes of two rfaC⁺ strains, E. coli R477 (Fig. 8, lane 2) and D21 (Fig. 9, lane 3), can catalyze ADP-mannose (but not GDP-mannose) dependent modification of Kdo₂-[4-³²P]-lipid IV_A, even though ADP-mannose is not known to exist in cells. To determine if this reaction of ADP-mannose and Kdo₂-[4-³²P]-lipid IV_A is catalyzed by heptosyl transferase I (20), E. coli strains deficient in heptosyl transferase I were examined. A role for heptosyl transferase I (RfaC) must be considered given the structural similarity of ADP-mannose and ADP-L-glycero-D-manno-heptose (Fig. 2), the presumed physiological donor of heptose residues in E. coli (20, 21, 22, 23). Membranes of strain D21f2 (20), which lacks a functional heptosyl transferase I due to a point mutation in rfaC, are unable to carry out the ADP-mannose dependent reaction (Fig. 9, lane 5). When strain D21f2 is transformed with plasmid pKZ84 (20) that contains the rfaC⁺ gene of S. typhimurium, the ability to utilize ADP-mannose is restored (Fig. 9, lane 7). At the dilution of the membranes employed, the observed shift of Kdo₂-[4-³²P]-lipid IV_A is entirely dependent upon the inclusion of exogenous ADP-mannose in each instance and is not the result of endogenous sugar donors present in the extracts (Fig. 9, lanes 2, 4, and 6).

Specific Activity of Kdo₂[4-³²P]-lipid IV_A Glycosylation using ADP- or GDP-mannose in Various Extracts

The mannosyl transferase of Rhizobium and the heptosyl transferase I of E. coli were measured under standard assay conditions that were linear with time and protein (Fig. 4), using Kdo₂-[4-³²P]-lipid IV_A as the acceptor. As seen in Table I, the mannosyl transferase of R. leguminosarum catalyzes the reaction about 2-fold more rapidly with GDP-mannose than with ADP-mannose. In the E. coli strains tested, there is no measurable activity with GDP-mannose, but ADP-mannose supports the glycosylation of Kdo₂-[4-³²P]-lipid IV_A in all the rfaC⁺ E. coli strains examined (Table I). The specific activities of the E. coli extracts are only 2-4-fold lower than those of R. leguminosarum when assayed with ADP-mannose. As expected from the qualitative data in Fig. 9, the heptosyl transferase I-deficient mutant, D21f2, displays no measurable activity with either sugar nucleotide. However, the specific activity of heptosyl transferase I in extracts of D21f2 transformed with pKZ84 is increased 150-fold over the wild type parental level observed in extracts of strain D21.

Table I. Comparison of mannosyl transferase specific activities with GDP- or ADP-mannose in extracts of selected strains Crude extracts were assayed under standard conditions using 1 mM sugar nucleotide and 6.25 µM Kdo2-[4-32P]-lipid IVA as substrates under standard conditions at 30 °C. Strain Specific activity (pmol/min/mg) GDP-mannose ADP-mannose R. leguminosarum 24AR 263 168 E. coli R477 (rfaC+) 0 92.5 E. coli D21 (rfaC+) 0 44.0 E. coli D21f2 (rfaC) 0 0 E. coli D21f2/pKZ84 0 6.51 × 103

DISCUSSION

Previous studies have demonstrated that R. leguminosarum contains the same seven enzymes found in E. coli to generate the LPS precursor, Kdo₂-lipid IV_A, starting with UDP-GlcNAc, R-3-hydroxymyristoyl-ACP, CMP-Kdo, and ATP (13). Each bacterial species then processes the Kdo₂-lipid IV_A in distinct ways to make its own type of lipid A and core (1). For example, E. coli extracts can add two heptose residues to Kdo₂-lipid IV_A (20), while R. leguminosarum extracts incorporate mannose and galactose (Fig. 1) (18).

In the present work we have established the first quantitative enzymatic assay for inner core assembly in R. leguminosarum. Furthermore, mannose transfer to Kdo₂-lipid IV_A (Fig. 3 and 4) has not been reported previously in any other system. In future studies with purified mannosyl transferase, it will be necessary to identify the site of attachment and linkage of the mannose residue that has been incorporated. As discussed below, we favor the view that mannose is attached via an 1-5-linkage to the inner Kdo residue, but this has not been demonstrated directly with the product generated in vitro in R. leguminosarum extracts. Degradation and sugar analysis of the enzymatic product does reveal the presence of a terminal mannose in the product,² consistent with the radiochemical experiments (Figs. 3 and 4). Not enough intact compound was generated to allow detection by fast atom bombardment mass spectrometry. The well behaved character of the mannosyl transferase assay (Figs. 4 and 5) will facilitate purification, molecular cloning, and overexpression of the R. leguminosarum mannosyl transferase, as well as allowing synthesis of larger amounts of product.

Lipid IV_A, which lacks the Kdo domain, cannot serve as a substrate for the R. leguminosarum mannosyl transferase reaction (Fig. 7). This is consistent with the proposed linkage of mannose to Kdo (Fig. 1). We might have expected to see mannosyl transferase activity with Kdo-lipid IV_A as the substrate, since mannose might be covalently linked to the Kdo present in that substrate. However, there was no evidence of such reaction (Fig. 7). This does not prove or disprove the possibility of an 1-5-linkage to the inner Kdo of Kdo₂-lipid IV_A. It may simply indicate that the single Kdo moiety of Kdo-lipid IV_A is not efficiently recognized by the R. leguminosarum mannosyl transferase. A striking dependence on both Kdo residues is also observed with the lauroyl transferase of E. coli encoded by the htrB gene (37, 38) that acts on Kdo₂-lipid IV_A.

The finding that ADP-mannose appears to substitute for GDP-mannose in extracts of R. leguminosarum (Fig. 8) may be due to the fact that adenine and guanine have similar structures. However, it is uncertain whether the same enzyme actually utilizes both nucleotides, or whether selective isoenzymes are responsible. Given that there is no known physiological role for ADP-mannose (a synthetic analog of GDP-mannose), we favor the view that a single mannosyl transferase uses both substrates. To settle the issue, the mannosyl transferase(s) will have to be purified and cloned.

The fact that GDP-mannose is not recognized as a sugar donor for the glycosylation of Kdo₂-lipid IV_A in cell extracts of E. coli (Fig. 3 and Table I) is consistent with the fact that mannose is not present in the inner core of E. coli LPS (1, 3). The finding that ADP-mannose does appear to be a substrate for the in vitro glycosylation of Kdo₂-lipid IV_A in E. coli extracts is explained by its structural resemblance to ADP-L-glycero-D-manno-heptose (Fig. 2). The absence of mannose from the E. coli inner core (1, 3) provides additional support for the view that ADP-mannose is a non-physiological analog that is not produced in vivo.

Until now, no quantitative enzymatic assay had existed for the E. coli heptosyl transferase I due to the inaccessibility of ADP-heptose (1, 20, 23). The generation of mannosyl-Kdo₂-lipid IV_A by the action of heptosyl transferase I in E. coli extracts is possible with ADP-mannose as an alternative substrate (Figs. 8 and 9). With this discovery, careful characterization and purification of heptosyl transferase I are now possible. The absence of mannosyl transferase activity in extracts of rfaC mutants conclusively proves that heptosyl transferase I is responsible for the observed activity in the E. coli system. Furthermore, it is extremely likely that mannose is added to the inner Kdo in this case via an 1-5-linkage, as indicated in Fig. 2. Linkage analysis of the mannose residues incorporated by the E. coli and R. leguminosarum enzymes will have to be carried out in parallel to determine whether or not the same structures are indeed generated.

In E. coli and S. typhimurium, it is possible to construct viable mutants lacking heptose (1, 3, 6). Such "deep rough" mutants are hypersensitive to antibiotics and detergents (1, 3, 6, 40). They are also deficient in certain outer membrane porins, the assembly and stability of which is influenced by the composition of the outer membrane lipopolysaccharide (3, 7, 41, 42, 43). It will be interesting to determine whether or not R. leguminosarum can grow without mannose in its inner core. This possibility is not altogether unlikely, since mutants lacking galactose (Fig. 1) have already been reported (15, 24).

It will be especially interesting to express the R. leguminosarum mannosyl transferase in E. coli mutants defective in the rfaC gene (1, 3, 20), and also engineered to express an rfb gene for GDP-mannose pyrophosphorylase (1, 39). Such strains might be able to generate a complete lipopolysaccharide, since GDP-mannose and ADP-heptose would both be available. Presumably the R. leguminosarum mannosyl transferase would add mannose to the inner Kdo, and RfaF might be able to utilize this mannose to incorporate what is normally the outer heptose. The approach of expressing unique R. leguminosarum genes in certain E. coli strains should enable the generation of new LPS structures in living cells and facilitate studies of LPS function.