Lipopolysaccharide biosynthesis in Rhizobium leguminosarum. Novel enzymes that process precursors containing 3-deoxy-D-manno-octulosonic acid.

The lipopolysaccharide of Rhizobium leguminosarum differs from that of other Gram-negative organisms. R. leguminosarum lipid A lacks phosphate groups, but it contains a galacturonic acid residue at the 4′-position and an aminogluconate moiety in place of the usual glucosamine 1-phosphate unit. R. leguminosarum lipid A is esterified with a peculiar long chain fatty acid, 27-hydroxyoctacosanoate, not found in enteric Gram-negative bacteria, and the inner core of R. leguminosarum contains mannose and galactose in place of heptose. Despite these differences, the biosynthesis of R. leguminosarum lipid A is initiated by the same seven enzyme pathway as in Escherichia coli (Raetz, C. R. H. (1993) J. Bacteriol.175, 5745-5753) to form the phosphorylated precursor, (Kdo)2-lipid IVA, which is then processed differently. We now describe several novel Rhizobium-specific enzymes that recognize and modify (Kdo)2-lipid IVA. The 1- and 4′-phosphatases were detected using (Kdo)2-[1-32P]-lipid IVA and (Kdo)2-[4′-32P]-lipid IVA, respectively, as shown by release of 32Pi. In the presence of GDP-mannose and/or UDP-galactose, membranes of R. leguminosarum first transferred mannose and then galactose to (Kdo)2-[4′-32P]-lipid IVA. In addition, at least two hydrophobic metabolites were generated from (Kdo)2-[4′-32P]-lipid IVA in a manner that was dependent upon both membranes and a cytosolic factor from R. leguminosarum. These compounds are attributed to novel acylations of (Kdo)2-[4′-32P]-lipid IVA. E. coli membranes and cytosol did not catalyze any of the unique reactions detected in R. leguminosarum extracts. Our findings establish the conservation and versatility of (Kdo)2-lipid IVA as a lipid A precursor in bacteria.

Lipopolysaccharides (LPSs), 1 or endotoxins, comprise the outer leaflet of the outer membranes of Gram-negative bacteria (1)(2)(3)(4)(5)(6)(7). LPS consists of three covalently linked domains. These are lipid A, the hydrophobic portion of the molecule that anchors LPS in the membrane, a core oligosaccharide consisting of inner and outer regions, and a repeating O-antigen. Biosynthesis of the lipid A portion is essential for cell viability (2,3,8,9). In addition, lipid A is responsible for the toxic effects observed when LPS is introduced into the mammalian bloodstream (3, 5, 10 -12). The endotoxin activity of lipid A results from the overproduction of cytokines by the immune system in response to lipid A (3, 5, 10 -12). These biological effects of Escherichia coli lipid A (Fig. 1) require the presence of several key structural features: both phosphate groups, the glucosamine disaccharide, and all the fatty acyl chains, especially the acyloxyacyl residues (3,5,10,12).
Although lipid A varies slightly in structure among different animal pathogens (1,13), the above hallmark structural features are generally conserved. However, the lipid A from the nitrogen-fixing bacterium Rhizobium leguminosarum differs strikingly from that of E. coli (14). It lacks the phosphate groups altogether ( Fig. 1) (14). Instead, it contains a galacturonic acid residue in place of the 4Ј-phosphate and an aminogluconic acid moiety in place of the proximal glucosamine 1-phosphate ( Fig. 1) (14). Preliminary structural studies suggest that it does not possess any acyloxyacyl residues (14), but instead, contains an unusual very long fatty acid, 27-hydroxyoctacosanoic acid (15). R. leguminosarum lipid A therefore lacks all the key structural features thought to be necessary for immunostimulation and toxicity in mammals (1,3,5,10). As yet, the immunostimulatory properties of R. leguminosarum lipid A have not been explored.
In addition to its distinct lipid A moiety, R. leguminosarum also possesses a different core structure (Fig. 2) (16). While the Kdo region is likely to be identical to that of E. coli LPS (17), the inner core of R. leguminosarum contains mannose instead of heptose, as well as galactose and galacturonic acid residues ( Fig. 2) (16).
Despite these marked differences in its LPS structure, R. leguminosarum extracts nevertheless contain all seven enzymes previously identified in E. coli for the biosynthesis of (Kdo) 2 -lipid IV A , an important precursor of LPS ( Fig. 3) (17). The fact that R. leguminosarum and E. coli make the same intermediate en route to LPS suggests that R. leguminosarum contains other enzymes, not present in E. coli, for the further processing of (Kdo) 2 -lipid IV A to its own unique lipid A. We have previously reported a 4Ј-phosphatase activity in extracts of R. leguminosarum that acts on (Kdo) 2 -lipid IV A (18). We now present evidence for a 1-phosphatase, a mannosyl transferase, a galactosyl transferase, and a long chain acyl transferase. All these transformations of (Kdo) 2 -lipid IV A occur in extracts of R. leguminosarum, but not of E. coli. The availability of these new enzymes will facilitate the preparation of novel endotoxin analogs for studies of lipid A activation of animal cells.
viciae 8401 was obtained from J. A. Downie (John Innes Institute, Norwich, United Kingdom). R. leguminosarum biovar phaseoli CE3 was a gift of D. Noel (Marquette University, Milwaukee, WI) (19). This strain was recently reclassified as Rhizobium etli (20). Rhizobium meliloti 1021 was obtained from Sharon Long (Stanford University). E. coli K12 strain R477 has been used for previously studies of lipid A biosynthesis in our laboratory (21). All rhizobia were grown on TY medium (17), containing 5 g of tryptone and 3 g of yeast extract per liter, and 10 mM CaCl 2 . Rhizobia were selected with 20 g/ml nalidixic acid and 200 g/ml streptomycin, when appropriate. All strains were grown at 30°C.
Preparation of Cell-free Extracts-Bacterial cultures were harvested in late logarithmic phase (A 550 ϭ 0.6 -1.0) by centrifugation at 8000 ϫ g av for 15 min, and the cell pellet was resuspended in 50 mM HEPES, pH 7.5, to give a final protein concentration of 5-15 mg/ml. The cells were broken by passage through a French pressure cell at 18,000 p.s.i. Unbroken cells and debris were removed by another centrifugation at 8000 ϫ g av for 15 min. Extracts were prepared and handled at 0 -4°C. Protein concentrations were determined with bicinchoninic acid (22), using bovine serum albumin for the standard curve. Subcellular fractions were prepared by centrifugation of the crude extract in 25 mM HEPES, pH 7.5, at 150,000 ϫ g av for 60 min. The supernatant was removed by aspiration. The supernatant was centrifuged a second time, and the second small pellet of residual membranes was discarded. The first membrane pellet was resuspended in the original volume of buffer. The resuspended membranes were centrifuged a second time, and the buffer wash was discarded. The final washed membranes were resuspended in ϳ20% of the original volume of crude extract from which they were made. The final preparations are referred to as the "cytosol" and the "washed membranes" in subsequent experiments.
Assays of (Kdo) 2 -lipid IV A Metabolism-Reaction mixtures contained 50 mM MES, pH 6.5, 0.5% Triton X-100, 10 mM dithiothreitol, 10 M (Kdo) 2 -[4Ј-32 P]-lipid IV A or (Kdo) 2 -[1-32 P]-lipid IV A (20,000 -50,000 cpm/ nmol), and crude cell-free extracts or subcellular fractions, as indicated, in a 10-l final volume. The reactions were incubated for 60 min at 30°C, and 5-l samples were withdrawn and applied to thin-layer chromatography plates. After drying the spots with a cool air stream, the plates were developed in CHCl 3 , pyridine, 88% formic acid, water (30:70:16:10, v/v). Exposure to imaging screens at room temperature or to x-ray film at Ϫ80°C was carried out overnight. Extent of conversion of substrate to product(s) was measured using a Molecular Dynamics PhosphorImager operated with ImageQuant software.
Size Fractionation of the Cytosol-A membrane-free cytosol was prepared as described above, except that during growth the cells were supplemented with 0.2% glucose and harvested at A 550 of approximately 2.5. This was done to increase the number of cells, but it had no effect on the biochemical activities present in the cytosol. Approximately 10 ml of cytosol (12.1 mg/ml protein) was applied to a total of eight Centricon C100 units. The units were centrifuged for about 4 h at 1000 ϫ g av at 4°C, according to manufacturer's directions, until 7.8 ml (containing 2.0 mg/ml protein) had emerged. Theoretically, this fraction of the cytosol should contain proteins less than 100,000 daltons in size, and is referred to as "C100 filtrate." About 7.0 ml of this material was applied to four Centricon C50 units, which were then centrifuged at 5000 ϫ g av for 15 min. The "C50 retentate" (1.0 ml, 6.9 mg), which should contain proteins with native molecular masses between 50 and 100 kDa, was removed and stored in aliquots at Ϫ80°C. The retentate is concentrated about 7 times on the basis of volume, as compared to the initial cytosol. Small molecules should not have become concentrated in the retentate, but should be at the same concentration as in the initial cytosol. The C50 filtrate (5.8 ml, 4.1 mg) presumably contains proteins less than 50 kDa in size, as well as other small molecules.

Rationale for Probing the Metabolism of (Kdo) 2 -lipid IV
acceptor for both the lauroyl and the myristoyl moieties that are present in acyloxyacyl linkage ( Fig. 1), as well as for the heptose residues of the inner core ( Fig. 2) (25)(26)(27). Since (Kdo) 2 -lipid IV A is also generated in extracts of Rhizobium (17), it is a plausible substrate for several Rhizobium-specific lipid A modifications (Figs. 1-3). One of these Rhizobium-specific reactions, a 4Ј-phosphatase (18), has already been identified. Here, we investigate whether or not other Rhizobium-specific reactions might exist that recognize (Kdo) 2 -lipid IV A as a substrate (Fig. 3).
Conversion of (Kdo) 2 -lipid IV A to Several Novel Products with or without GDP-Mannose and UDP-Galactose-In the absence of added sugar nucleotides, (Kdo) 2 -[4Ј-32 P]-lipid IV A is rapidly converted to several products when incubated with crude R. leguminosarum extracts. A major product is 32 P i , which is produced by the 4Ј-phosphatase (Fig. 4, lane 2) (18). The lipid product generated by the 4Ј-phosphatase from (Kdo) 2 -[4Ј-32 P]lipid IV A is unlabeled, and therefore it is not seen in Fig. 4. However, under the conditions employed, several more hydrophobic bands are also observed (Fig. 4, lane 2). As shown below, we believe that the band labeled "a and a" is a mixture of (Kdo) 2 -[4Ј-32 P]-lipid IV A derivatives that either are further acylated (metabolite a) or are dephosphorylated at the 1-position (metabolite a). These can be resolved from each other by  (14). Distinct features of the proposed structure of R. leguminosarum lipid A, as compared to that of E. coli, are the lack of phosphate residues at positions 1 and 4Ј, the presence of galacturonic acid at position 4Ј, the presence of 2-aminogluconic acid in place of the proximal glucosamine 1-phosphate unit, and the presence of an unusual very long chain fatty acid, 27-hydroxyoctacosanoate. The latter is apparently ester linked, but it is not part of an acyloxyacyl unit (14). Laurate and myristate are absent in R. leguminosarum lipid A. modifying the conditions of thin layer chromatography. Small amounts of an additional, very hydrophobic compound designated a؆ can also be seen (Fig. 4, lane 2).
To look for core sugar extensions beyond Kdo, (Kdo) 2 -[4Ј-32 P]-lipid IV A was incubated with crude extracts of R. leguminosarum 8401 and the potential sugar donors, GDP-mannose and/or UDP-galactose (Fig. 4, lanes 3-5). In the presence of GDP-mannose, both the (Kdo) 2 -[4Ј-32 P]-lipid IV A and its more hydrophobic derivative(s) shifted to more slowly-migrating positions (Fig. 4, lane 3, metabolites b and b), consistent with the addition of a mannose residue. In the presence of UDP-galactose, no additional metabolites of (Kdo) 2 -[4Ј-32 P]lipid IV A , beyond those formed by the crude extract alone, were observed (Fig. 4, lane 4 compared to lane 2). When both sugar nucleotide donors were present, however, an additional, even more slowly migrating derivative of (Kdo) 2 -[4Ј-32 P]-lipid IV A was generated (Fig. 4, lane 5, metabolite c).
We postulate that metabolite b represents the addition of mannose to (Kdo) 2 -[4Ј-32 P]-lipid IV A , and c represents the further addition of galactose to mannosyl-(Kdo) 2 -[4Ј-32 P]-lipid IV A . These results support the proposed core structure of R. leguminosarum LPS (Fig. 2), according to which galactose transfer would only be possible after mannose is added (16). When non-radioactive (Kdo) 2 -lipid IV A is used as the substrate together with GDP-[U-14 C]mannose and/or UDP-[U-14 C]galactose, the presence of 14 C is detected only in those additional bands generated in the presence of each sugar nucleotide (data not shown). A similar pattern of glycosylations of (Kdo) 2 -[4Ј-32 P]-lipid IV A is seen in extracts of R. leguminosarum CE3 (data not shown).
Direct Evidence for 1-Phosphatase Activity-The complex reactions occurring in crude extracts of R. leguminosarum were further investigated by using (Kdo) 2 -[1-32 P]-lipid IV A as the probe (Fig. 5). A phosphatase acting on the 1-position, if present, would release 32 P i from (Kdo) 2 -[1-32 P]-lipid IV A . As shown in Fig. 5, 32 P i generation was indeed catalyzed by R. leguminosarum extracts (lane 2). The 1-phosphatase was stimulated severalfold by the inclusion of 1 mM ATP (Fig. 5, lane 3). The 1-phosphatase was not selective for substrates bearing the Kdo domain, since it also dephosphorylated earlier precursors in the lipid A pathway, such as lipid IV A and lipid X (1-3) (data not shown).
An additional advantage of using (Kdo) 2 -[1-32 P]-lipid IV A as the substrate is that the lipid product generated by the 4Јphosphatase reaction can now be visualized. Thus, the prominent, more rapidly migrating derivative of (Kdo) 2 -[1-32 P]-lipid IV A observed in lane 2 of Fig. 5 (metabolite p) mainly reflects the action of the 4Ј-phosphatase (18). As noted above, metabolite p cannot be detected in the experiment of Fig. 4, since p is not radioactive after (Kdo) 2 -[4Ј-32 P]-lipid IV A has been dephosphorylated at position 4Ј. Metabolites a and p are distinct monodephosphorylated isomers of (Kdo) 2 -lipid IV A that happen to migrate with about the same R F . A small amout of metabolite a also presumably migrates with p in Fig. 5, lane 2.
In the presence of GDP-mannose, or GDP-mannose plus UDP-galactose (Fig. 5, lanes 4 and 5), several more slowly migrating metabolites are formed from (Kdo) 2 -[1-32 P]-lipid IV A (designated q, b, and c), similar to what is seen with (Kdo) 2 -[4Ј-32 P]-lipid IV A (Fig. 4). When only UDP-galactose is added as the co-substrate with (Kdo) 2 (17). Heptose (L-glycero-D-manno-heptose) is absent in the proposed structure of the R. leguminosarum core, and it is replaced with mannose and galactose (16). In addition, the R. leguminosarum inner core is rich in galacturonic acid (16), which is absent in E. coli. Additional sub-stoichiometric modifications of the inner core of E. coli with phosphate, ethanolamine phosphates, or additional sugars (3,5) are not indicated.

FIG. 3. Conserved biosynthesis followed by divergent processing of (Kdo) 2 -lipid IV A in extracts of R. leguminosarum and E.
coli. The same seven enzymes are present to generate (Kdo) 2 -lipid IV A in extracts of both organisms (17,18). The 4Ј-phosphate, which is labeled with 32 P in many of the experiments described in the current work, is indicated with an arrow. Shown below the structure of (Kdo) 2lipid IV A are the diverse enzymes identified so far that act on (Kdo) 2lipid IV A in extracts of R. leguminosarum, compared to those previously identified in extracts of E. coli (25)(26)(27).
tive of (Kdo) 2 -[1-32 P]-lipid IV A (metabolite b in Fig. 5), when isolated, is a substrate for the 4Ј-phosphatase (data not shown), which accounts for the appearance of metabolite q in the presence of GDP-mannose (Fig. 5, lanes 4 and 5). In addition, some q may be generated by the action of the mannosyl transferase on 4Ј-dephospho-(Kdo) 2 -[1-32 P]-lipid IV A (metabolite p). Like p, metabolite q cannot be visualized when (Kdo) 2 -[4Ј-32 P]-lipid IV A is used as the subtrate (Fig. 4), since q has lost its 4Јphosphate residue.
Core Sugar Additions in Other Rhizobium Species-While R. leguminosarum lipid A lacks phosphate (Fig. 1), lipid A of R. meliloti appears to be more like that of E. coli in that the usual two phosphates are present (28,29). 2 In R. meliloti extracts, there is no evidence for 4Ј-and 1-phosphatase activities (18). Although the core structure of R. meliloti has not been reported, preliminary studies of its composition indicate that mannose is absent, but that galactose and glucose are present (28,29). 2 We therefore wanted to know if extracts of R. meliloti contained the same or different core glycosyl transferases as R. leguminosarum. The glycosylation of (Kdo) 2 -[4Ј-32 P]-lipid IV A was examined, using extracts of R. meliloti 1021 (data not shown). No glycosylation products of (Kdo) 2 -[4Ј-32 P]-lipid IV A were generated in the presence of GDP-mannose alone, but significant glycosylation was observed with UDP-galactose alone (data not shown). We conclude that the enzymatic transfers of inner core sugars onto (Kdo) 2 -[4Ј-32 P]-lipid IV A in extracts of R. meliloti differ significantly from those observed in extracts of R. leguminosarum, consistent with the limited information that is available regarding the core domain of R. meliloti.
The mannosyl and galactosyl transferase activities, like the phosphatases, appeared in the washed membranes (Fig. 6, lanes 2-5), but not in the cytosol (Fig. 6, lanes 6 -9). When washed membranes and cytosol were recombined (Fig. 6, lanes  10 -13), however, significantly larger amounts of several more hydrophobic derivatives of (Kdo) 2 -[4Ј-32 P]-lipid IV A were generated than when (Kdo) 2 -[4Ј-32 P]-lipid IV A was incubated with membranes alone. This result is seen most clearly in Fig. 6, lanes 14 -19, in the absence of added sugar nucleotides. When both membranes and cytosol were used (Fig. 6, lanes 18 and  19), there was a substantial increase in the amounts of at least two more hydrophobic metabolites (a/a and a؆) of (Kdo) 2 -[4Ј-32 P]-lipid IV A , as compared to assays in which membranes alone or cytosol alone were used (Fig. 6, lanes 14 -17). The most hydrophobic of these metabolites (a؆) was not detected at all in the incubations containing membranes or cytosol alone (Fig. 6,  lanes 1-9). Its formation was also stimulated severalfold by the presence of 1 mM ATP (Fig. 6, lane 19).
Size Fractionation of the Cytosolic Factor Required for the Generation of the Hydrophobic Derivatives-We wanted to investigate further the formation of the more hydrophobic metabolites of (Kdo) 2 -[4Ј-32 P]-lipid IV A involving the interaction of the membranes and the cytosol of R. leguminosarum (Fig. 6). The whole cytosol was first fractionated by centrifugation through a 100-kDa sizing membrane, as described under "Experimental Procedures." The filtrate was then centrifuged through a 50-kDa membrane, concentrating the soluble macromolecules that range in size from approximately 50 to 100 kDa by about 7-fold relative to the unfractionated cytosol.
Next, (Kdo) 2 -[4Ј-32 P]-lipid IV A was incubated with washed membranes and various size-fractionated pools of the cytosol (Fig. 7). In this reconstituted system, in which more extensive conversion of substrate to hydrophobic products was possible, one could discern two closely migrating substances just above (Kdo) 2 -[4Ј-32 P]-lipid IV A (designated a and a), and one sub-2 R. Carlson, personal communication.

FIG. 5. Metabolism of (Kdo) 2 -[1-32 P]-lipid IV A stimulated by ATP, GDP-mannose, and UDP-galactose in extracts of R. leguminosarum.
Assay conditions were similar to those shown in Fig. 4, except that 10 M (Kdo) 2 -[1-32 P]-lipid IV A was used as the probe. Crude extract of R. leguminosarum 8401 was present at 3.6 mg/ml, and ATP, GDP-mannose, and UDP-galactose were added at 1.0 mM, as indicated. stantially higher migrating product (a؆). In the earlier chromatograms (for instance in Figs. 4 and 6), these two closely migrating compounds (a and a) had not been resolved from one another. Separation was enhanced in the experiment of Fig. 7 by using chromatography solvent that was 2-3 days old and by developing the chromatogram all the way to the top of the plate. Thus, we found that what appeared initially as a single band in Figs. 4 and 6 was actually a combination of two substances (a and a in Fig. 7), which were produced in varying amounts depending on the exact assay conditions.
Membranes alone (Fig. 7, lane 2) produced only product a from (Kdo) 2 -[4Ј-32 P]-lipid IV A . Whole cytosol alone was inactive, as were the two size-fractionated pools of the cytosol (Fig.  7, lanes 3-5). However, when combined, the washed mem-branes and the cytosol produced all three hydrophobic metabolites (a, a, and a؆ in Fig. 7, lane 6). Metabolites a and a؆ only appeared in the presence of both washed membranes and cytosol. The active component of the cytosol ran through a 100-kDa molecular filtration device. However, it was retained and concentrated by a 50-kDa filtration membrane (Fig. 7, lanes 6  and 7 versus lane 8).
We suggest that metabolite a in Fig. 7 is a derivative of (Kdo) 2 -[4Ј-32 P]-lipid IV A that is dephosphorylated at position 1. Metabolite a contains a distinct modification that makes it more hyrdophobic than (Kdo) 2  metabolite a reflects a novel acylation of (Kdo) 2 -[4Ј-32 P]-lipid IV A with a long fatty acid. 3 Mild alkaline hydrolysis of metabolite a with 0.2 M NaOH results in the formation of a deacylation product that is the same as the one that is obtained by alkaline hydrolysis of (Kdo) 2 -[4Ј-32 P]-lipid IV A . This observation is consistent with the presence of an additional esterlinked acyl chain in metabolite a. Furthermore, metabolite a migrates significantly faster than monolauroyl-(Kdo) 2 -[4Ј-32 P]lipid IV A generated by the E. coli Kdo-dependent acyltransferase, HtrB (26). Therefore, we believe that the R. leguminosarum membrane-bound component may be a novel, long chain acyltransferase, and the cytosolic factor is a unique, macromolecular long chain acyl donor. 3 Following this reasoning, metabolite a؆ in Fig. 7 might be (Kdo) 2 -[4Ј-32 P]-lipid IV A that is both dephosphorylated at the 1-position and further acylated with a long chain fatty acid.

Transformations of (Kdo) 2 -[4Ј-32 P]-lipid IV A Observed in Extracts of R. leguminosarum Do Not Occur in E. coli Extracts-
To determine if the reactions observed in R. leguminosarum extracts are also present in E. coli, (Kdo) 2 -[4Ј-32 P]-lipid IV A was incubated with extracts of E. coli using the exact same assay conditions developed above for R. leguminosarum (Fig.  8). A typical pattern of metabolites, both with and without ATP, GDP-mannose, and UDP-galactose, was observed with R. leguminosarum (Fig. 8, lanes 3 and 4), and all of the (Kdo) 2 -[4Ј-32 P]-lipid IV A was consumed in the complete system (Fig. 8,  lane 3). In contrast, no derivatives of (Kdo) 2 -[4Ј-32 P]-lipid IV A were observed when matched E. coli extracts were substituted (lane 2). These results establish that the metabolites described here represent novel enzymatic transformations of (Kdo) 2 -[4Ј-32 P]-lipid IV A that could not have been observed previously using the E. coli system (2,3). DISCUSSION An overview of the proposed reactions that account for the processing of (Kdo) 2 -lipid IV A observed in extracts of R. leguminosarum is shown in Figs. 3 and 9. In addition to the 4Јphosphatase (18), the present work provides evidence for the following Rhizobium-specific enzymes: 1) a mannosyl and a galactosyl transferase that act sequentially on (Kdo) 2 -lipid IV A (Figs. 4, 5, and 6). The mannose is presumably added to one of the Kdo residues, as lipid IV A is not a mannose acceptor. The mannosyl-(Kdo) 2 -lipid IV A can still be dephosphorylated after mannose addition (Fig. 9). 2) A phosphatase that acts at the 1-position of (Kdo) 2 -lipid IV A (Fig. 9), as shown by inorganic phosphate release from (Kdo) 2 -[1-32 P]-lipid IV A (Fig. 5). In crude extracts, the 1-phosphatase is stimulated severalfold by 1 mM ATP (Fig. 5). However, the significance of this observation cannot be judged until the 1-phosphatase is purified. 3) A novel acylation of (Kdo) 2 -lipid IV A (Figs. 7 and 9, metabolite a), which depends on a membrane component and a cytosolic factor (Fig. 7). This reaction represents the acylation of (Kdo) 2 -[4Ј-32 P]-lipid IV A with 27-hydroxyoctacosanoic acid, a fatty acid that is unique to Rhizobium (15). 3 Lastly, we propose that the most rapidly migrating product, metabolite a؆ in Fig. 7, is a derivative of (Kdo) 2 -[4Ј-32 P]-lipid IV A that is both dephosphorylated at the 1-position and acylated with a long chain fatty acid (Fig. 9). E. coli membranes and cytosol (Fig. 8) do not support the generation of any of the metabolites shown in Fig. 9.
With the discovery of these unique, efficient transformations of (Kdo) 2 -lipid IV A , the stage is set for the purification of the novel enzymes that catalyze them. This effort will facilitate the unambiguous identification of the multiple products that are being generated by the crude R. leguminosarum extracts that we have employed. Even if our schematic structural proposals (Fig. 9) are not correct in all details, we have clearly established the fact that (Kdo) 2 -lipid IV A is recognized by a variety of distinct new enzymes, further supporting the view that (Kdo) 2 - . The utilization of (Kdo) 2 -lipid IV A in R. leguminosarum is catalyzed by at least four enzymes (the 1-and 4Ј-phosphatases, the mannosyl transferase, and the putative long chain acyltransferase) that can work in parallel. This processing results in many different metabolites, the major forms of which are indicated, and the identification of which is based on the results with the radiolabeled (Kdo) 2 -lipid IV A probes used in the experiments of Figs. 3-8. Additional metabolites must exist that we cannot yet detect, given the presence of galacturonic acid residues and aminogluconate in the mature lipid A of R. leguminosarum (Fig. 1) (14). In future studies, the above metabolites will have to be isolated in larger amounts to permit unambiguous assignment of their structures. lipid IV A is a key, generally conserved intermediate of LPS assembly in diverse microbial systems.
Several additional enzymes besides the ones that we have identified (Figs. 3 and 9) must exist in R. leguminosarum to generate its unique lipid A. For instance, there must be an enzyme that transfers galacturonic acid to the 4Ј-position after dephosphorylation. So far, attempts to demonstrate galacturonic acid transfer from UDP-galacturonic acid to 4Ј-dephosphorylated (Kdo) 2 -lipid IV A (metabolite p of Fig. 9) have not been successful. There must also be enzymatic mechanisms for the oxidation of the 1-position of the lipid A disaccharide (Fig.  1) to generate the aminogluconate moiety. Attempts to demonstrate oxidation following 1-dephosphorylation of (Kdo) 2 -lipid IV A (i.e. of metabolite a of Fig. 9) have not yielded positive results. An enzymatic precedent for the oxidation of the anomeric carbon at the reducing end of an oligosaccharide has been reported in the case of cellobiose dehydrogenase (30).
To study the relationship between the function of Rhizobium LPS and its special structural features, the isolation of mutants in the new enzymes that we have discovered will be required. Such mutants might have a lipopolysaccharide structure more closely resembling that of E. coli. The reasons for the special set of enzymes that we have found in Rhizobium might become clear by characterizing such mutants. Special structural modifications of lipid A might be required for root hair infection, nodule formation, or maintenance of symbiosis (31). Perhaps, some plants can mount an "immune" response to bacteria containing phosphorylated lipid A residues.
In parallel with purification and mutant isolation, we intend to clone the genes encoding the enzymes that we have discovered and to express them in E. coli. In this way it may be possible to modify the structure of lipid A in living cells of E. coli. It will be very interesting to determine whether or not E. coli can grow with lipid A having the unusual structural features associated with R. leguminosarum. The approach of modifying lipid A structure in living cells may provide new insights into the biological functions of lipid A and the assembly of outer membranes.