An Outer Membrane Enzyme That Generates the 2-Amino-2- deoxy-gluconate Moiety of Rhizobium leguminosarum Lipid A* 210

The structures of Rhizobium leguminosarum and Rhizobium etli lipid A are distinct from those found in other Gram-negative bacteria. Whereas the more typical Escherichia coli lipid A is a hexa-acylated disaccharide of glucosamine that is phosphorylated at positions 1 and 4′, R. etli and R. leguminosarum lipid A consists of a mixture of structurally related species (designated A–E) that lack phosphate. A conserved distal unit, comprised of a diacylated glucosamine moiety with galacturonic acid residue at position 4′ and a secondary 27-hydroxyoctacosanoyl (27-OH-C28) as part of a 2′ acyloxyacyl moiety, is present in all five components. The proximal end is heterogeneous, differing in the number and lengths of acyl chains and in the identity of the sugar itself. A proximal glucosamine unit is present in B and C, but an unusual 2-amino-2-deoxy-gluconate moiety is found in D-1 and E. We now demonstrate that membranes ofR. leguminosarum and R. etli can convert B to D-1 in a reaction that requires added detergent and is inhibited by EDTA. Membranes of Sinorhizobium meliloti and E. coli lack this activity. Mass spectrometry demonstrates that B is oxidized in vitro to a substance that is 16 atomic mass units larger, consistent with the formation of D-1. The oxidation of the lipid A proximal unit is also demonstrated by matrix-assisted laser desorption ionization time-of-flight mass spectrometry in the positive and negative modes using the model substrate, 1-dephospho-lipid IVA. With this material, an additional intermediate (or by product) is detected that is tentatively identified as a lactone derivative of 1-dephospho-lipid IVA. The enzyme, presumed to be an oxidase, is located exclusively in the outer membrane of R. leguminosarum as judged by sucrose gradient analysis. To our knowledge, an oxidase associated with the outer membranes of Gram-negative bacteria has not been reported previously.

The Rhizobiacea are agriculturally important Gram-negative bacteria that are able to establish a symbiotic relationship with the root cells of certain plants (1). The symbiotic bacteria provide the plant with NH 4 ϩ by fixing N 2 , whereas the plant provides the bacteria with reduced carbon sources (2). A de-tailed understanding of the many factors contributing to the complex interplay between the plant host and the microbes is beginning to emerge (3). A few of the key components needed for effective symbiosis and nitrogen fixation include flavonoids (4), Nod factors (5,6), receptor kinases (7), various exopolysaccharides (8,9), cyclic ␤-glucans (10,11), K-antigens (12), and lipopolysaccharides (LPSs) 1 (13,14).
LPSs are macromolecular glycolipids present on the outer surfaces of the outer membranes of Gram-negative bacteria (15)(16)(17)(18). The structure of LPS may be subdivided into the lipid A region, which embeds the LPS in the outer membrane, the nonrepeating core oligosaccharide, and the distal O-antigen polysaccharide. The effects of changing LPS structure on symbiosis and nitrogen fixation are not fully characterized (14). Mutants of Rhizobium leguminosarum and Rhizobium etli that lack O-antigen are defective in the infection process, producing poorly differentiated, nonfunctional nodules (13, 19 -21). Whether or not mutations in lipid A biosynthesis are deleterious to symbiosis remains to be determined.
Recent studies of the LPS of R. leguminosarum and R. etli, which are distinct bacterial species based on their ribosomal RNA sequences (22), have revealed that both bacteria possess unusual lipid A molecules (23)(24)(25) and core oligosaccharides (14,20,26,27). Although the classical lipid A structure, typified by that of Escherichia coli, consists of a hexa-acylated disaccharide of glucosamine that is phosphorylated at positions 1 and 4Ј (15)(16)(17)(18), R. etli and R. leguminosarum lipid A is recovered as a mixture of structurally related components (A to E) that are not phosphorylated ( Fig. 1) (24,25,28). A conserved distal unit, comprised of a diacylated glucosamine residue with a secondary 27-OH-C28 acyl chain as part of an acyloxyacyl moiety at position 2Ј and a galacturonic acid residue at position 4Ј, is present in all of the components ( Fig. 1) (24,25,28). The microheterogeneity of R. etli lipid A arises mainly in the proximal unit, which may contain acyl chains of different lengths at position 2 or may be deacylated at position 3 ( Fig. 1). Although glucosamine constitutes the proximal units of components B and C, its oxidized form, 2-amino-2-deoxy-gluconate (2-aminogluconate), is present in D-1 and E ( Fig. 1) (24,25,28). Components C and E are 3-O-deacylated derivatives of the more abundant species B and D-1, respectively (24,25,28). Component A (not shown in Fig. 1) appears to be an elimination product generated from D-1 during the mild acid hydrolysis procedure used to release lipid A from the LPS core (24,25). D-2 is an isomer formed from D-1 in a nonenzymatic reaction and is attributed to acyl chain migration on the proximal unit (24,25).
To date, the occurrence of the 2-aminogluconate moiety is limited to the lipid A of R. etli and R. leguminosarum (23,24). A plausible pathway for 2-aminogluconate formation could involve the oxidation of the proximal glucosamine unit of component B to generate D-1. In conjunction with our recent reevaluation of the structures of R. etli and R. leguminosarum lipid A (25), we discovered that crude extracts of R. etli and R. leguminosarum can convert 14 C-labeled component B to a lipid that migrates with D-1 during thin layer chromatography. We now report the detailed characterization of this novel oxidative reaction using both the natural product B and the relatively simple model substrate, 1-dephospho-lipid IV A . Remarkably, the enzyme is an outer membrane protein and is presumed to be an oxidase. These findings are consistent with the idea that 2-aminogluconate formation is a late step in R. leguminosarum lipid A biosynthesis. As demonstrated in the accompanying manuscript, a novel gene, designated lpxQ, encodes the putative oxidase.

EXPERIMENTAL PROCEDURES
Materials-Glass-backed 0.25-mm Silica Gel 60 thin layer chromatography plates were from Merck. Chloroform, ammonium acetate, and sodium acetate were obtained from EM Science. Pyridine, methanol, and formic acid were from Mallinckrodt. [U-14 C]acetate was purchased from Amersham Biosciences.
Bacterial Growth Conditions and Membrane Preparations-R. leguminosarum 3855 was grown at 30°C in TY broth (5 g of tryptone and 3 g of yeast extract/liter) supplemented with 10 mM CaCl 2 . R. etli CE3 and Sinorhizobium meliloti 1021 were grown in TY broth supplemented with 10 mM CaCl 2 , 20 g/ml nalidixic acid, and 200 g/ml streptomycin sulfate. E. coli strains were grown at 37°C in LB broth (29) with one of the following antibiotics, depending on the resistance markers of the plasmid that the strain harbors: ampicillin (100 g/ml), tetracycline (15 g/ml), and kanamycin (25 g/ml). Table I describes the various bacterial strains used.
Rhizobium cultures (0.5 to 1 liter) were grown to late log phase (OD 550 ϭ ϳ1.0) and then harvested by centrifugation at 6,000 ϫ g for 15 min at 4°C. All subsequent steps were carried out at 4°C. The cell pellets were washed with 50 mM HEPES, pH 7.5, with a volume that was 1/10 th that of the original culture volume. The cells were collected by centrifugation and resuspended in a minimal volume (usually 5 ml) of 50 mM HEPES, pH 7.5, and stored at Ϫ80°C if they were not immediately lysed. The frozen cells were thawed, resuspended in 50 mM HEPES, pH 7.5, at a protein concentration of ϳ3-10 mg/ml, and broken by three passages through an ice-cold French pressure cell (SLM Instruments, Urbana, IL) at 18,000 p.s.i. Large cell debris and unbroken cells were removed by centrifugation at 12,100 ϫ g for 10 min. The membranes were recovered from the cytosol by ultracentrifugation at 149,000 ϫ g for 1 h. The pellet containing the membranes was washed by homogenization in the same volume of 50 mM HEPES, pH 7.5. After a second ultracentrifugation, the washed membranes were homogenized in 50 mM HEPES, pH 7.5, at a protein concentration of ϳ5-15 mg/ml. The samples were stored in aliquots at Ϫ80°C. The protein concentrations were determined by the bicinchoninic acid assay with bovine serum albumin as the standard (30).
Subcellular Localization of the Lipid A Oxidase-Subcellular localization of the lipid A oxidase in R. leguminosarum 3855 membranes was determined using a protocol similar to that described by Trent et al. (31). Briefly, a 1-liter culture of R. leguminosarum 3855 was grown with shaking (225 rpm) at 30°C for 16 h (OD 550 ϭ ϳ1.0). The culture was centrifuged at 6,000 ϫ g for 10 min at 4°C. The cell pellet was resuspended in 7.8 ml of 50 mM HEPES, pH 7.5, containing 0.5 mM EDTA, and the cells were lysed by passage through a French pressure cell at 10,000 p.s.i. Crude cell-free extract was prepared by removal of unbroken cells and large debris by centrifugation at 12,100 ϫ g for 10 min at 4°C, and the supernatant was recovered. One-half of the crude extract was stored at Ϫ80°C, whereas the remaining half was used to prepare washed membranes, as described above. The washed membranes were homogenized with a 25-gauge 1/2 syringe needle in a total volume of 2.5 ml of 50 mM HEPES, pH 7.5, containing 0.5 mM EDTA. After being layered on top of a seven-step isopycnic sucrose gradient as described by Guy-Caffey et al. (32), the membranes were centrifuged in a swinging bucket rotor at 155,000 ϫ g for 18 h at 4°C. The fractions (0.5. ml) were collected, and their protein content was determined by the bicinchoninic acid assay, as described above. The following undiluted fraction volumes were used for the various assays: 3 l for the lipid A oxidase reaction, which was terminated after 30 min; 50 l for the NADH oxidase assay (33,34); and 20 l for the phospholipase A assay (35). The turbidity (OD 600 ) of each fraction was measured to confirm the presence of membrane fragments. The activity for each fraction was calculated as a percentage of the total activity throughout the entire gradient.
FIG. 1. Structures of the predominant lipid A species in E. coli, R. etli, and R. leguminosarum. R. etli lipid A, which is thought to be the same as that of R. leguminosarum, is a mixture of several related species (23)(24)(25). In contrast to E. coli lipid A, all R. etli lipid A species lack phosphate groups (24,25). Instead, each one contains a galacturonic acid moiety at position 4Ј and a single acyloxyacyl unit, featuring an unusual 27-OH-C28 secondary acyl chain, at position 2Ј (24,25). The dashed bonds indicate partial substitutions in the major R. etli components B and D-1 (24,25). The molecular weight of the largest, fully substituted form of each component is indicated in parentheses. The minor R. etli lipid A species C and E (not shown) are the 3-O-deacylated derivatives of B and D-1, respectively (24,25). The proximal sugar is a glucosamine unit in B and an aminogluconate moiety in D-1 (24,25). The proximal 3-deoxy-D-manno-2-octulosonic acid (Kdo) residue of the core oligosaccharide (not present in lipid A prepared by mild acid hydrolysis) is attached at position 6Ј in intact LPS.
Preparation of Nonradioactive and 14 C-Labeled R. etli Lipid A Component B-Nonradioactive B was prepared from R. etli CE3 as described previously (24,25). To make 14 C-labeled component B, a 50-ml culture of R. etli CE3 was grown to A 550 ϭ 1 in the presence of 500 Ci of [U-14 C]-acetate (50 mCi/mmol) (24,25). A combination of DEAE-cellulose column chromatography and preparative thin layer chromatography (24) was used to purify [ 14 C]B.
Preparation of 1-Dephospho-lipid IV A -Lipid IV A was isolated as described by Raetz et al. (36) and further purified by reversed phase column chromatography (37). To cleave the phosphate at the anomeric position (38, 39), 3.2 mg of lipid IV A in a 16 ϫ 125-mm glass tube with a Teflon-lined cap was resuspended by sonic irradiation in 3.6 ml of 0.1 M HCl and placed in a boiling water bath for 15 min. After cooling, the hydrolyzed material was converted into a two-phase Bligh-Dyer system (40) by the addition of 4 ml of CHCl 3 and 4 ml of MeOH. After mixing, the solution was centrifuged in a clinical centrifuge for 10 min. The lower phase was removed and transferred to a clean glass tube. The upper phase was re-extracted once with 4 ml of a fresh lower phase of a pre-equilibrated two-phase Bligh-Dyer system (40). The lower phases from both extractions were pooled and dried down under a stream of N 2 . The dried product was then redissolved in 450 l of CHCl 3 /MeOH (4:1, v/v) and spotted in a line 20 mm from the edge of a 20 ϫ 20-cm silica TLC plate. The plate was allowed to dry and developed in CHCl 3 , pyridine, 88% formic acid, H 2 O (50:50:16:5, v/v) until the solvent front was ϳ5 cm from the top of the plate. The solvents were removed by drying the plate with a cold air stream for 30 min. The band of interest, which is transiently visible as a white zone during the drying process, was localized by placing the plate on top of a light box. The band was outlined in pencil and scraped off with a clean razor blade. The silica chips were transferred into a 16 ϫ 125-mm glass tube equipped with a Teflon-lined cap. The lipid was extracted from the chips by resuspending them in 3.8 ml of an acidic single-phase Bligh-Dyer solution, consisting of CHCl 3 , MeOH, 0.1 M HCl (1:2:0.8, v/v), followed by brief sonic irradiation. The solution was then converted into a two-phase Bligh-Dyer mixture by adding 3 ml of CHCl 3 , 2 ml of MeOH, and 2.8 ml of distilled H 2 O. After thorough mixing and centrifugation for 8 min in a clinical centrifuge, the lower phase was recovered and passed through a Pasteur pipette plugged with glass wool. The extraction of the chips was repeated once, and three drops of pyridine were added to the pooled lower phases to neutralize any residual HCl. After drying the lower phases with a stream of N 2 , the sample was stored at Ϫ20°C.
Further purification of the 1-dephospho-lipid IV A was done by passing the material from the preparative TLC step through a 0.5-ml DEAE-cellulose column (Whatman DE-52), previously equilibrated as the acetate form in CHCl 3 , MeOH, H 2 O (2:3:1, v/v) (24,36,41). The above sample was redissolved in 12 ml of CHCl 3 , MeOH, H 2 O (2:3:1, v/v) and loaded onto the column. The flow-through was collected as one fraction. The column was washed with 1 ml of CHCl 3 , MeOH, H 2 O (2:3:1, v/v), and stepwise elution was achieved by including increasing ammonium acetate concentrations in the solvent mixture (24,36). The column was thus successively washed with 2 ml of CHCl 3 , MeOH, 30 mM aqueous NH 4 Ac (2:3:1, v/v), 2 ml of CHCl 3 , MeOH, 60 mM aqueous NH 4 Ac (2:3:1, v/v), 4 ml of CHCl 3 , MeOH, 120 mM aqueous NH 4 Ac (2:3:1, v/v), 2 ml of CHCl 3 , MeOH, 240 mM aqueous NH 4 Ac (2:3:1, v/v), and finally with 2 ml of CHCl 3 , MeOH, 500 mM aqueous NH 4 Ac (2:3:1, v/v). 0.5-ml fractions were collected throughout. Elution of the lipid from the column was monitored by TLC using silica gel plates (5 ϫ 10 cm), which were developed with the solvent CHCl 3 , pyridine, 88% formic acid, H 2 O (50:50:16:5, v/v). Some of the 1-dephospho-lipid IV A emerged from the column in the late CHCl 3 , MeOH, 60 mM NH 4 Ac wash, but most of the material eluted in the CHCl 3 , MeOH, 120 mM NH 4 Ac step. Only the fractions from the 120 mM wash were processed further as the source of 1-dephospho-lipid IV A substrate for in vitro incubations with the R. leguminosarum membranes. The selected fractions were pooled and converted into a two-phase Bligh-Dyer system, consisting of CHCl 3 , MeOH, H 2 O (2:2:1.8, v/v). The solvents were mixed and centrifuged for 8 min at room temperature in a clinical centrifuge. The upper phase was re-extracted once with a pre-equilibrated lower phase. The desired lower phases were then pooled and dried with a stream of N 2 . The purification yielded about 1.0 mg of 1-dephospho-lipid IV A .
The 1 H NMR analysis (2D COSY) of the final compound redissolved in 600 l of CDCl 3 , CD 3 OD, D 2 O (2:3:1, v/v) and analyzed under the conditions described previously (42) confirmed its identity as 1-dephospho-lipid IV A (supplementary figure). The spectrum of the purified 1-dephospho-lipid IV A , which is referenced against internal tetramethylsilane, indicates that about 90% of the sample adopts the ␣-anomeric configuration in the proximal glucosamine unit, but the ␤-anomer and its cross-peaks are also detected. Following NMR spectroscopy, the 1-dephospho-lipid IV A was dried down under N 2 and resuspended in 750 l of 50 mM MOPS, pH 7.0, to make a 1 mM aqueous dispersion. This was stored at Ϫ80°C and subjected to brief sonic irradiation prior to use in assays.

Preparation of [4Ј-32 P]1-dephospho-lipid IV A -
The starting material [4Ј-32 P]lipid IV A was synthesized enzymatically as previously described (43) and then was converted into [4Ј-32 P]1-dephospho-lipid IV A by hydrolysis in 0.1 M HCl. Typically, a sample containing ϳ80 -100 Ci of [4Ј-32 P]lipid IV A was re-suspended in 150 l of 0.1 M HCl. Following sonic irradiation for 2 min in a bath apparatus, the mixture was placed in a 100°C heat block for 15 min. After cooling, the hydrolyzed material was centrifuged briefly and then spotted onto a 20 ϫ 20-cm Silica gel plate. Preparative TLC was carried out as described above for the nonradiolabeled lipid IV A . After drying, the product was detected by a brief autoradiography and recovered by extraction from the silica chips (see above). The final substrate was typically dispersed in 50 mM HEPES, pH 7.5, such that the working stock solution contained ϳ50,000 cpm/l.

Assay Conditions for Measuring the Conversion of [ 14 C]B to [ 14 C]D-1-
The standard reaction mixture (10 l) contained 10 M [ 14 C]B (ϳ500 cpm/reaction), 0.5-1.0 mg/ml membrane protein, 0.1% Triton X-100, 1 mM MgCl 2 , and 50 mM MES buffer, pH 6.5, unless otherwise indicated. The reactions were incubated under ambient conditions at 30°C and terminated at the indicated times by spotting 4-l samples onto a 20 ϫ 20-cm silica gel TLC plate. The spots were dried for 30 min with a cold air stream, and the plate was then developed in the solvent CHCl 3 , MeOH, H 2 O/pyridine (40:25:4:2, v/v). The remaining substrate and product(s) were detected with a Molecular Dynamics Storm PhosphorImager equipped with ImageQuant software. Enzyme specific activity (usually expressed as nmol/min/mg) was calculated based on the percentage of conversion of substrate to product(s).  the initial and final time points indicated complete conversion of B to a compound migrating with a component D-1 standard. Half of the reaction mixture (contained in a sterile 16 ϫ 125-mm glass tube) was transferred to an identical glass tube. Each half was then converted into a two-phase Bligh-Dyer mixture by the addition of CHCl 3 (4 ml) and MeOH (4 ml) to each tube. After mixing, the phases were separated by centrifugation for 10 min at room temperature in a clinical centrifuge. The lower phase was recovered and transferred to a clean glass tube. After two extractions of the upper phases with pre-equilibrated lower phases, all of the resulting lower phases were pooled, dried under N 2 , and stored at Ϫ20°C.  (24). Lipids in fractions from the flow-through, 30 mM NH 4 Ac, and 60 mM NH 4 Ac washes were pooled separately and recovered by two-phase Bligh-Dyer partitioning, as discussed above for the preparation of 1-dephospholipid IV A . The lower phases for each pool were dried down with a stream of N 2 and stored at Ϫ20°C.

Assay Conditions for Detecting the Oxidation of [4Ј-32 P]1-Dephospholipid IV
Isolation of the Products Generated in Vitro from 1-Dephospho-lipid IV A by R. leguminosarum Membranes-In a sterile 16 ϫ 125-mm glass tube equipped with Teflon-lined cap, 100 M 1-dephospho-lipid IV A was incubated with 0.36 mg/ml R. leguminosarum 3855 membranes in a buffer containing 50 mM MOPS, pH 7.0, 0.1% Triton X-100, and 1 mM MgCl 2 . The final reaction volume was 7 ml. A parallel reaction in a 1.5-ml microcentrifuge tube containing 96 l of the above reaction mixture and 4 l of [4Ј-32 P]lipid IV A ϳ(50,000 cpm/l) was carried out to monitor product formation with a PhosphorImager.
After 14 h at 30°C, half of the nonradioactive reaction mixture was transferred to a second identical glass tube. CHCl 3 (4 ml), MeOH (4 ml), and 0.1 M HCl (360 l) were added to each of the two tubes, which were mixed and centrifuged for 10 min in a clinical centrifuge. After two more extractions of the upper phase with pre-equilibrated lower phase, all of the lower phases were pooled in a clean glass tube. Two drops of pyridine were added before the pooled lower phases were dried with N 2 and stored at Ϫ20°C. Fractions from each elution step were pooled separately and recovered by two-phase Bligh-Dyer partitioning as described above for 1-dephospho-lipid IV A . The desired lower phases were dried down, and the lipids were stored at Ϫ20°C prior to further analysis.
Mass Spectrometry-Matrix-assisted laser desorption ionization timeof-flight (MALDI/TOF) mass spectra were acquired on a Kompact MALDI 4 from Kratos Analytical (Manchester, UK), equipped with a nitrogen laser (337 nm), 20 kV extraction voltage, and time delayed extraction. The samples were prepared for MALDI/TOF analysis by depositing 0.3 l of the lipid sample dissolved in chloroform/methanol (4:1, v/v), followed by 0.3 l of a saturated solution of 2,5-dihydroxybenzoic acid in 50% acetonitrile as the matrix. The samples were left to dry at room temperature before the spectra were acquired in both the positive and negative ion linear modes. Each spectrum was the average of 50 laser shots. Fig. 2, the membranes of R. leguminosarum 3855 efficiently convert [ 14 C]B to a substance migrating like [ 14 C]D-1 in the presence of Triton X-100. The cytosolic fraction is completely inactive (Fig. 2), and it does not further stimulate the activity present in membranes (not shown). Other nonionic detergents such as Nonidet P-40 can substitute for Triton X-100, but ionic detergents, like SDS or LDAO, are inhibitory (not shown). Product formation is nearly linear with time for about 30 min at 0.25 mg/ml membrane protein (Fig. 3). The enzyme is inhibited by addition of 5 mM EDTA to the assay mixture (Fig. 4A, lane 2a). However, nearly complete reactivation of the EDTA-inhibited enzyme is observed upon addition of excess (10 mM) MgCl 2 and incubation for another 20 min (Fig. 4B, lane 7). Co 2ϩ , Ni 2ϩ , and Mn 2ϩ are comparable with Mg 2ϩ in reactivating the EDTA-treated enzyme (Fig. 4B). Ca 2ϩ , Zn 2ϩ , Cu 2ϩ , and Fe 2ϩ are minimally effective, but Mo 2ϩ is inactive (Fig. 4B). The chelators dipyridyl and o-phenanthroline did not inhibit the reaction when added at 10 mM.

Conversion of [ 14 C]B to [ 14 C]D-1 by R. leguminosarum and R. etli Membranes-As shown in
The membranes of other strains of R. leguminosarum or R. etli catalyze the conversion of [ 14 C]B to [ 14 C]D-1 at rates that are comparable with those seen with membranes of R. leguminosarum 3855 (Fig. 5). The membranes of E. coli and S. meliloti 1021, which make lipid A species that are fully phosphorylated and do not contain 2-aminogluconate, are unable to metabolize [ 14 C]B to [ 14 C]D-1 (Fig. 5).
A scaled-up, nonradioactive enzymatic reaction mixture was used to prepare D-1 for mass spectrometry. Following purification on a DEAE-cellulose column, MALDI/TOF mass spectra of the remaining substrate B (Fig. 6, lower panel) and of the D-1-like material (Fig. 6, upper panel) were recorded in the positive reflectron mode. As noted in Fig. 1, both the substrate and the product are mixtures of related lipid species differing in acyl chain length by two methylene groups (i.e. 28 atomic mass units) on the proximal unit (24,25). The observed peaks at m/z 1980.1 and 2008.1 (Fig. 6, lower panel) may be interpreted as the monoisotopic [M ϩ Na] ϩ ions for the two predominant acyl chain lengths present in B (Fig. 1) (24). The putative D-1 generated in vitro by R. leguminosarum membrane gives strong peaks at m/z 1996.5 and 2024.5 (Fig. 6, upper panel), consistent with the incorporation of a single oxygen atom by oxidation of the proximal glucosamine unit to the 2-aminogluconate residue (24). As with B, both signals may be interpreted as the monoisotopic [M ϩ Na] ϩ ions arising from the two predominant acyl chain lengths present in D-1 (Fig. 1). The distributions of the additional peaks at progressively higher m/z in each of the clusters shown in Fig. 6 are consistent with the molecular weights of B and D-1 (Fig. 1) and are in agreement with previous studies of the natural products isolated from R. etli CE3 using conventional MALDI/TOF mass spectrometry (24). The additional microheterogeneity of B and D-1 because of the partial substitution with ␤-hydroxybutyrate ( Fig. 1) (24) is not seen in the region of the spectrum shown in Fig. 6.

Use of [4Ј-32 P]1-Dephospho-lipid IV A as an Alternative Substrate for Demonstrating Oxidation of the Proximal Glucosamine Unit-The E. coli lipid A precursor [4Ј-32 P]lipid IV A ,
which is a tetra-acylated disaccharide of glucosamine that is phosphorylated at positions 1 and 4Ј (36,44), is not metabolized appreciably by R. leguminosarum membranes under the oxidase assay conditions (Fig. 7A, lower arrow). A 1-phosphatase is present in R. leguminosarum that converts R. leguminosarum lipid A precursors to their 1-dephosphorylated derivatives (45), but this activity is barely detectible with [4Ј-32 P]lipid IV A as the substrate under the conditions employed. However, when [4Ј-32 P]1-dephospho-lipid IV A (Fig. 8 and supplementary  figure) is prepared by chemical hydrolysis of lipid IV A with 0.1 M HCl at 100°C and used as a substrate under the oxidase assay conditions, it is metabolized to several new products ( 7B). These compounds migrate slightly faster than the substrate in this solvent system (Fig. 7B). Almost all of the 1-dephospho-lipid IV A is consumed after overnight incubation at 30°C, as shown by mass spectrometry (see below). The same lipid A oxidizing enzyme that converts B to D-1 is responsible for the metabolism of 1-dephospho-lipid IV A by R. leguminosarum membranes (89).
The substrate 1-dephospho-lipid IV A (Fig. 9A) shows the expected peak at m/z 1324.8, which is interpreted as [M Ϫ H] Ϫ . The proposed 2-aminogluconate product (Fig. 9C) displays an intense peak at m/z 1341.2, consistent with [M Ϫ H] Ϫ of a compound that has gained an oxygen atom (Fig. 8). The peak at m/z 928.9 (Fig. 9C) could not be assigned and might be an impurity carried over from the R. leguminosarum membranes used as the enzyme source.
The negative mode MALDI/TOF mass spectrum of the second (less abundant) product derived from 1-dephospho-lipid IV A , which emerges from DEAE-cellulose with the 60 mM NH 4 Ac wash, is shown in Fig. 9B. An intense signal is observed at m/z 1078.9. The putative lactone derivative of 1-dephospholipid IV A (Fig. 8) has a molecular weight of 1323.7 and should give rise to a peak at m/z 1322.7 in the negative mode. How-ever, such lactones might undergo elimination of the fatty acid moiety at position 3 (24,25). In the case of the putative lactone derived from 1-dephospho-lipid IV A , the elimination of hydroxymyristic acid would generate an ␣/␤-unsaturated lactone (Fig. 8) with a molecular weight of 1079.3. Consequently, the major peak seen at m/z 1078.9 in Fig. 9B could be interpreted as [M Ϫ H] Ϫ derived from the lactone elimination product (Fig.  8). We do not believe that the putative elimination product forms prior to mass spectrometry, because it would migrate more slowly during TLC than 1-dephospho-lipid IV A , which is not observed (Fig. 7B).
A small amount of residual substrate is evident in the sample shown in Fig. 9B, consistent with its charge (Fig. 8), as judged by the peak at m/z 1325.2. As in Fig. 9C, the peak  Fig. 1 (24). The upper panel shows the same region of the partial spectrum of D-1 synthesized in vitro, which consists of molecules that are all 16 mass units larger than their counterparts in B. The peaks at m/z 1996.5 and 2024.5 correspond to the expected monoisotopic [M ϩ Na] ϩ ions for the two major forms of D-1, which differ by two methylene groups (28 atomic mass units), consistent with the analysis of the same lipid A molecules isolated from cells, as reported by Que et al. (24). at m/z 928.9 (Fig. 9B) is not assigned and may represent an impurity.
Positive Ion MALDI/TOF Mass Spectrometry of Products Made by R. leguminosarum Membranes from 1-Dephospholipid IV A -The positive ion MALDI/TOF mass spectra (Fig. 10) of the same three substances are consistent with the interpretation of their negative mode spectra (Fig. 9). Most significantly, the positive mode spectra reveal an intense B 1 ϩ ion at m/z 695.3 in the substrate (Fig. 10A), the proposed 2-aminogluconate (Fig. 10C), and the putative lactone elimination product (Fig. 10B). The presence of a common B 1 ϩ ion clearly demonstrates that the distal units of these lipids are identical and supports the view that the proximal sugar unit of 1-dephospholipid IV A is converted to the 2-aminogluconate derivative (Fig.  8) during the course of the reaction shown in Fig. 7. However, these findings do not clearly distinguish the aldehyde oxidation pathway (Fig. 8, reaction 1) from the alternative of a lactone intermediate (Fig. 8, reactions 2 and 3), because the lactone might also be formed from the 2-aminogluconate product of reaction 1 by a nonenzymatic process (Fig. 8, reaction 4).
Outer Membrane Localization of the Lipid A Oxidase of R. leguminosarum-In previous studies, we have described methods for separating inner and outer membranes of various Rhizobium strains using isopycnic sucrose gradient centrifugation (46). As in E. coli, the heavier (outer) membrane fraction is characterized by its phospholipase A activity, whereas the lighter (inner) membranes are detected using NADH oxidase. Remarkably, the lipid A oxidase activity, which was measured by following the conversion of [ 14 C]B to [ 14 C]D-1, is recovered almost entirely with the outer membrane fragments of R. leguminosarum 3855 (Fig. 11). The same profile was obtained when 1-dephospho-lipid IV A was used as the substrate (not shown). Among the various R. etli or R. leguminosarum lipid A enzymes studied to date (45)(46)(47)(48)(49)(50)(51), only the oxidase is found in the outer membrane. The outer membrane localization of the oxidase therefore demonstrates that the formation of 2-aminogluconate represents a late modification in the maturation of R. leguminosarum lipid A. DISCUSSION R. leguminosarum and R. etli are the only organisms known to synthesize 2-aminogluconate, an oxidized derivative of Dglucosamine (23)(24)(25)52). This material substitutes for the proximal glucosamine unit that is usually present in lipid A (Fig. 1)  (23-25, 52). In re-evaluating the microheterogeneity and structures of R. leguminosarum and R. etli lipid A, Que et al. ( 2 and 3), a lactone intermediate is generated by oxidation of the proximal pyranose of 1-dephospho-lipid IV A , followed by lactonase catalyzed hydrolysis to generate the 2-aminogluconate unit. It is conceivable that the oxidation and the lactone opening are both catalyzed by the same protein or that the latter step is nonenzymatic. Although the intact lactone is not detected during mass spectrometry (see below), a fragment with the molecular weight expected for the proposed elimination product of the lactone is in fact very prominent. Furthermore, it also remains a formal possibility that the lactone is formed nonenzymatically from the 2-aminogluconate moiety (reaction 4). In the in vitro conversion of component B to D-1 (Figs. 1 to 6), the putative lactone derivative apparently does not accumulate to the extent seen with 1-dephospho-lipid IV A . 28) discovered that about one-third of the lipid A isolated from these organisms does in fact contain the conventional glucosamine disaccharide backbone found in most other Gram-negative bacteria (Fig. 1, structure B). In earlier studies by Bhat et al. (23), B had been overlooked, because the lipid A released from lipopolysaccharide by acid hydrolysis was not further purified. Given that B lacks the anomeric phosphate residue (24,25), it might be the immediate precursor of component D-1 (Fig. 1). We have now developed a quantitative assay for following the conversion of B to D-1 using membranes of R. leguminosarum or R. etli  and have demonstrated the presence of an additional oxygen atom in the D-1 product generated in vitro from B. We also report the model substrate, 1-dephospho-lipid IV A (Figs. 7 and 8), with which to characterize the enzyme.
A remarkable feature of the oxidase is its outer membrane localization. This finding suggests that the formation of 2-aminogluconate occurs as a late modification of the lipid A molecule. All of the early conserved reactions of lipid A biosynthesis in R. leguminosarum and R. etli, as well as the 4Ј and 1 phosphatases that are unique to these organisms, appear to be associated with the inner membrane (45)(46)(47)(48)(49)(50)(51). 2 Very few outer membrane enzymes have ever been described. All of the known outer membrane enzymes are lipases (35,53), acyltransferases (54,55), or proteases (56). The outer membrane localization of the enzyme that generates D-1 from B suggests that it is a novel kind of oxidase. Oxygen is indeed required for the reaction, as demonstrated in the accompanying manuscript. However, it is not yet possible to show stoichiometric formation of H 2 O 2 and D-1 from O 2 and B when using whole membrane preparations as the enzyme source (see below).
The ability of the oxidase to utilize 1-dephospho-lipid IV A (Figs. 7 and 8) 9. Negative-mode MALDI/TOF mass spectrometry of the oxidation products of 1-dephospho-lipid IV A . The lipids extracted from an overnight reaction mixture containing R. leguminosarum 3855 membranes and 1-dephospho-lipid IV A were fractionated on a DEAEcellulose column. A, the spectrum of the substrate 1-dephospho-lipid IV A prior to incubation with membranes. B, the lipids from the CHCl 3 , MeOH, 60 mM aqueous NH 4 Ac (2:3:1, v/v) wash, consisting of residual substrate and the putative lactone. C, the lipids from the CHCl 3 , MeOH, 120 mM aqueous NH 4 Ac (2:3:1, v/v) wash. The latter consist primarily of the proposed 2-aminogluconate derivative, as judged by the peak at m/z 1341.2. The peak at m/z 928.9 could not be assigned and may reflect contaminating R. leguminosarum membrane lipids.
FIG. 10. Positive-mode MALDI/TOF mass spectrometry of the oxidation products of 1-dephospho-lipid IV A . The lipids extracted from an overnight reaction mixture containing R. leguminosarum 3855 membranes and 1-dephospho-lipid IV A were fractionated on a DEAEcellulose column as in Fig. 9. A, the substrate 1-dephospho-lipid IV A . B, the lipids from the CHCl 3 , MeOH, 60 mM aqueous NH 4 Ac (2:3:1, v/v) wash. C, the lipids from the CHCl 3 , MeOH, 120 mM aqueous NH 4 Ac (2:3:1, v/v) wash. The peaks at m/z 931.1, 854.0, and 813.4 could not be assigned and may reflect contaminating R. leguminosarum membrane lipids. The B 1 ϩ ions of the substrate 1-dephospho-lipid IV A , the lactone elimination product, and the aminogluconate are all observed at m/z 695.3, demonstrating that the distal unit is unchanged and that only the proximal unit of the product is modified under these conditions. 1) for catalysis. However, the oxidase does require a free hydroxyl group at the anomeric position of the proximal lipid A unit, as shown by its inability to utilize lipid IV A (Fig. 7A). The oxidase also prefers a substrate with an acyl chain at position 3, given its relatively slow oxidation of component C (Fig. 1) compared with B (data not shown).
Two plausible routes for the oxidation of the proximal glucosamine unit are shown in Fig. 8. In the lactone pathway (Fig. 8,  reactions 2 and 3), oxidation occurs before ring opening, and a lactone intermediate is formed that could be hydrolyzed by a separate lactonase to generate the 2-aminogluconate moiety. Alternatively, the same protein might catalyze both the oxidation and lactone hydrolysis, or the latter step might even be nonenzymatic. In any case, an analogous lactone is synthesized by glucose oxidase and glucose dehydrogenase. Glucose oxidase from Aspergillus niger (57) is a soluble iron-dependent flavoenzyme (58) that converts glucose and oxygen to ␦-gluconolactone and H 2 O 2 . On the other hand, the glucose dehydrogenases are generally PQQ-dependent enzymes that likewise convert glucose to gluconolactone (59). Depending on the microorganism, glucose dehydrogenase can be membrane-bound or soluble and requires either Mg 2ϩ or Ca 2ϩ (59). Membrane-bound glucose dehydrogenase has been isolated from a wide range of bacteria (59). Its active site is presumed to be located on the periplasmic surface of the inner membrane, and the electron acceptor is ubiquinone (60, 61). A conserved region near the N terminus, predicted to consist of five membrane-spanning helices, is thought to be important for ubiquinone binding (62).
Although the membrane-bound glucose dehydrogenases oxidize various monosaccharides, they cannot process disaccharides. In contrast, an atypical soluble glucose dehydrogenase found only in Acinetobacter calcoaceticus (63,64) can oxidize both mono-and disaccharides. Interestingly, this periplasmic enzyme is rather different from the membrane-bound glucose dehydrogenases, not only with regard to its sugar substrate specificity but also in its preference for electron acceptors. It does not utilize ubiquinone (61) but instead slowly reduces a soluble cytochrome b that does not appear to interact with the electron transport chain (65). Furthermore, the periplasmic dehydrogenase shares very little amino acid sequence similarity with the membrane-bound enzymes (66), and it lacks the conserved 11-residue tryptophan docking motif found in most other PQQ-containing dehydrogenases (67).
Another enzyme that oxidizes the C1 of a hexose moiety is cellobiose dehydrogenase, which is found in the lignin-degrading white rot fungi Phanerochaete chrysosporium (68 -72) and Humicola insolens (73). Cellobiose dehydrogenase is an extracellular enzyme that oxidizes various di-and oligosaccharides and can utilize either Fe 3ϩ , O 2 , or various organic molecules as electron acceptors (74,75). As shown in the accompanying manuscript, the lipid A oxidase of R. leguminosarum does not share significant sequence homology with any of the above enzymes.
Another alternative for 2-aminogluconate formation might involve ring opening of the proximal lipid A unit, followed by oxidation the aldehyde (Fig. 8, reaction 1). This possibility is reminiscent of the mechanism proposed for D-xylose isomerase (76) (also known as D-glucose isomerase), which interconverts a broad spectrum of aldoses and ketoses (77). X-ray data suggest the presence of an extended open chain sugar substrate in the enzyme active site (78,79). Mg ϩ2 is believed to be the sole cofactor needed by the enzyme in vivo (80), but other divalent cations can be substituted in vitro.
Given the outer membrane localization of the lipid A oxidase, we consider it unlikely that its enzymatic mechanism involves pyridine or flavin nucleotides. This idea is supported by the observation that there is no stimulatory effect on the rate of conversion of B to D-1 by addition of exogenous NAD, NADP, FAD, FMN, ubiquinone, or cytochrome c (data not shown). Furthermore, the reaction is not inhibited by cyanide. A mechanism involving PQQ deserves consideration in view of the inner membrane or periplasmic localization of the PQQ-dependent glucose dehydrogenases discussed above, but again no stimulation of B to D-1 conversion was seen with added PQQ (data not shown). At present, the only clue to the mechanism of the oxidation is the inhibition by added EDTA and the reactivation with excess Mg 2ϩ and some other divalent cations, especially Co 2ϩ , Ni 2ϩ , and Mn 2ϩ (Fig. 4). Although a direct involvement of Mg 2ϩ in substrate binding, catalysis, or maintenance of tertiary structure is certainly a possibility, we cannot exclude the alternative that the enzyme requires a redoxactive heavy metal, which is removed by EDTA and then transferred back in the presence of excess Mg 2ϩ or other divalent cations (Fig. 4).
A mechanism in which electrons are transferred from B to molecular oxygen in the outer membrane without the involvement of the inner membrane electron transport chain seems attractive based upon the available data. This hypothesis predicts that H 2 O 2 would be formed as a by-product. Attempts to demonstrate stoichiometric formation of H 2 O 2 with the Amplex Red detection kit (Molecular Probes) during the conversion of 50 M B to D-1 have not been successful so far (data not shown). However, the membrane preparations used as the enzyme source rapidly consume 50 M H 2 O 2 added exogenously (data FIG. 11. Outer membrane localization of the lipid A oxidase in R. leguminosarum 3855. the washed membranes of R. leguminosarum 3855 were separated into inner and outer fractions by isopycnic sucrose density gradient centrifugation (46). Marker enzymes for the outer and inner membranes (phospholipase A2 and NADH oxidase, respectively) were used to confirm the separation. Top panel, turbidity (OD 600 ) was used to detect membrane fragments, and each fraction was also assayed for the R. leguminosarum lipid A oxidase at 30°C for 30 min under standard conditions by following the conversion of [ 14 C-]B to [ 14 C-]D-1. Lower panel, phospholipase A2 and NADH oxidase marker enzymes. not shown). Purification of the oxidase to homogeneity, as well as structural and mechanistic studies, will therefore be required to identify the intermediates and by-products of the conversion of B to D-1.
Although the occurrence of 2-aminogluconate is limited to a few Gram-negative bacteria involved in nitrogen fixation (23)(24)(25), an emerging general theme in lipid A biogenesis is that specific covalent modifications may occur within the outer membrane (31,54,55). In some cases, these outer membrane modifications are subject to exquisite transcriptional regulation (31,54). For instance, the lipid A acyltransferase PagP of E. coli and S. typhimurium is regulated by the PhoP/PhoQ two-component system (54), which is induced during growth on low concentrations of Mg 2ϩ (81,82). PagP is a typical outer membrane protein and is synthesized with a signal sequence that is cleaved during export (54). Recent structural NMR studies of PagP (55) have revealed that its active site faces the outside. It will be very interesting to determine the structure of the oxidase and establish whether or not it is regulated, especially during the differentiation of bacteroids to form nitrogenfixing nodules. Mutants lacking the lipid A oxidase will be necessary to study the function of the 2-aminogluconate residue. The identification of the structural gene encoding the oxidase, described in the accompanying article (89), should facilitate such further biological and chemical studies.